learn

short, practical guides on ai video, images, and signals — plus notes on how to use them well.

ai video & images

making cinematic clips with sound, animating stills, and what the other tools actually charge.

ai video & images · 4 min
AI lip sync video — one clip where it works, one where it doesn't Two real AI video clips with spoken lines, made the same week with the same character: one where the lip sync lands and one where it falls apart. What was asked, what came back, and what seems to make the difference.
ai video & images · 5 min
Free AI video generators in 2026 — what's actually free, and the catch Which AI video generators are really free in 2026, what each kind of free tier takes away (watermarks, commercial rights, resolution, queue time), which big assistants stopped making video, and the honest way to make a clip you can actually post.
ai video & images · 5 min
we did the math against runway and kling — and we lose on $/second We published the table with our loss in it. Runway and Kling beat us on dollars per second of video. Here is the math, and the four places the subscription number stops being true.
ai video & images · 5 min
the t-shirt and the car — what an ai video model quietly drops We spent a week getting three women, one t-shirt and a car into fifteen seconds. Here is exactly what the engine dropped at each step, why, and the two habits that stopped it.
ai video & images · 4 min
AI video with voiceover — what renders with the clip, and what does not A spoken line inside the shot and a narrator over the cut are made in different places. What renders with the clip, what has to be a separate track, and how to layer them without the result sounding pasted together.
ai video & images · 3 min
Can Copilot make videos? Not anymore — here's what happened Microsoft Copilot made video through OpenAI's Sora, in Bing Video Creator and then in Microsoft 365 Copilot Create. Sora was discontinued, Microsoft retired the feature in June 2026, and a replacement is "in development." What that leaves you with today.
ai video & images · 4 min
AI product video generator — what a short cinematic clip is actually for An honest guide to AI product video in 2026: what a short cinematic clip does for a listing, where template tools beat it, what to feed the model, and what a clip actually costs when you pay per render instead of per month.
ai video & images · 2 min
magic hour alternative — when you want one great clip, not a hundred tools Magic Hour bundles dozens of AI tools into a cheap subscription. When the toolbox is worth it, what the credit math hides, and the pay-per-clip alternative.
ai video & images · 3 min
chatbot video creator — direct a video by talking, not prompting What a chatbot video creator actually is, why composing a shot in conversation beats a prompt box, and how to direct an HD clip with sound by talking.
ai video & images · 3 min
can claude make videos? no — and here is what people do instead Claude still does not generate video — not on Fable 5.1, not on any plan. It can now output a video file from code, which is a different thing. What that distinction means, and where the actual rendering happens.
ai video & images · 2 min
can meta ai make videos? yes — and free is not the only price Yes — Meta AI generates video for free, but the generator moved into Vibes, its AI clip feed, and the product is built around sharing. What free actually costs you, and the pay-per-clip alternative.
ai video & images · 2 min
google flow alternative — when you want the clip, not the studio Google Flow is a filmmaking workspace bundled into Google AI plans, with credits that do not roll over. When it is worth it, and the pay-per-clip alternative.
ai video & images · 3 min
ai movie trailer — the one film form ai is actually built for A trailer is the one film form built from disconnected striking shots — which is exactly what AI video is good at. The shot list, how to handle the voice-over, the sound, and the real cost per clip.
ai video & images · 2 min
video-to-video vs image-to-video — which one do you want? Video-to-video restyles footage you already have. Image-to-video animates a still. Which one you actually want, and which tools do which.
ai video & images · 2 min
an ai video generator without credits — what you're really getting Tools advertising no credits still meter you — by minutes, renders, queue position or a monthly plan. What each model actually costs, and when credits are the honest option.
ai video & images · 2 min
storyboard to video with ai — one panel at a time Turning a storyboard into AI video means one clip per panel, not one prompt for the whole thing. How to shoot a board, what stays consistent, and what it costs.
ai video & images · 2 min
gemini omni alternative — pay per clip, no google account Gemini Omni replaced Veo in the Gemini app and needs a Google account and a plan. What it does, what it costs you in practice, and the pay-per-clip alternative.
ai video & images · 2 min
ai video for ads — promo clips without a production budget How to make short AI video ads that don't look generic — one product, one clear beat, native sound, and per-clip pricing so testing creative stays cheap.
ai video & images · 3 min
ai video extender — what you actually get when you extend a clip How AI video extenders work — chained blocks of a few seconds (about seven per pass on nixm), where quality starts drifting, and what each extension costs in credits.
ai video & images · 2 min
a2e alternative — what 30 free credits a day actually buys A2E gives 30 free credits a day, but image-to-video costs about 30 credits per clip and they expire nightly. The daily-credit math, and the pay-per-video alternative.
ai video & images · 2 min
open-source ai video — the real cost of running it yourself Wan, LTX, Hunyuan and the rest are free to download — the real cost is VRAM, setup and minutes per clip. What local AI video generation actually takes.
ai video & images · 2 min
Grok Imagine — video generation with a subscription attached Grok Imagine generates short AI video inside the Grok and X apps — gated behind X's subscription tiers. How it works, the subscription math, and the pay-per-video route.
ai video & images · 3 min
AI video for TikTok and Shorts — vertical, loud, and yours Making AI video for TikTok, Reels, and YouTube Shorts: the 9:16 workflow, why sound-in-one-pass matters, what stops the scroll, and the pay-per-clip economics of posting daily.
ai video & images · 2 min
How long can AI videos actually be in 2026? Typical AI video lengths in 2026: most models render 5-15 second clips, and longer pieces are edited sequences of short generations. Why the limits exist and how to work with them, not against them.
ai video & images · 2 min
"Unlimited" AI video — the honest arithmetic Thousands search for unlimited AI video generators. Here's the honest arithmetic: what "unlimited" means in practice, where the limits hide, and when pay-per-video beats every subscription.
ai video & images · 3 min
Bing Video Creator has shut down — here's where to go Microsoft shut down Bing Video Creator on July 31, 2026. What happens to your creations, why the free-Sora era ended, and the honest alternatives.
ai video & images · 3 min
Can ChatGPT make videos? Not anymore — here's what happened ChatGPT's video generator was Sora, and OpenAI has shut it down: the app in April 2026, the API in September. What that means if you are looking for an OpenAI video generator, and where chat-to-video lives now.
ai video & images · 2 min
Script to video — how a written page actually becomes footage How script-to-video AI actually works in 2026 — why a script becomes a shot list, not one long render, and how to film it scene by scene with native sound.
ai video & images · 2 min
Realistic AI video — what actually sells the shot Why some AI videos look real and most don't — the physics, light, and sound that sell realism, the tells that break it, and the workflow that gets you there.
ai video & images · 2 min
AI video without the watermark — what the stamp is really for Why free AI video tiers watermark everything, what removal really costs across tools, and the honest per-clip route to clean footage you own.
ai video & images · 2 min
4K AI video — the real numbers, and when you actually need it What 4K AI video generation actually involves in 2026 — the real render times, the credit multipliers across tools, and when HD is honestly the better buy.
ai video & images · 3 min
Can Gemini make videos? Yes — here's what stands between you and the render Yes. Gemini generates ~10-second clips with native audio through Gemini Omni (Veo 3.1 in Flow), but only on Google AI Plus, Pro or Ultra — not the free plan. What each tier unlocks, where the caps bite, and the per-clip alternative.
ai video & images · 3 min
NotebookLM video — a document explainer, not a film camera NotebookLM turns your documents into narrated explainer videos — sometimes taking 30+ minutes. What it's actually for, its limits, and when you want a cinematic tool instead.
ai video & images · 4 min
Higgsfield vs pay-as-you-go — one subscription, many exchange rates Higgsfield bundles 30+ video and image models behind one subscription, but every model spends credits at its own rate, credits reset each cycle, and "unlimited" has a speed governor. The honest per-clip maths, and when the bundle is worth it.
ai video & images · 6 min
How to make an AI video — the whole workflow, honestly How to make an AI video in 2026: describe one moment, compose the frame, render it with sound, then extend it or keep the same character across shots. The full workflow, the mistakes that waste credits, and what a finished piece costs.
ai video & images · 2 min
AI video for YouTube — Shorts, b-roll, and the honest workflow How to make AI video for YouTube in 2026 — Shorts built from one striking clip, cinematic b-roll for longer videos, paid per clip instead of per month.
ai video & images · 3 min
AI slop — why the first render looks wrong, and what to do about it Why your AI videos look wrong, and the workflow that fixes it: judge the frame before you animate, re-roll cheaply, keep the one take that lands. No subscription.
ai video & images · 6 min
Short film with AI — how to hold a story together across shots Make a short film with AI video in 2026: hold a character, a light and a grade across shots, plan coverage, and cut it together. The honest workflow, no subscription.
ai video & images · 7 min
Image to video AI — turn a photo into a clip, and what it costs Turn a photo or picture into an AI video clip with sound. What image-to-video actually does, a real still and the clip it became, what drives the cost, where the free options stop, and what a single clip runs.
ai video & images · 6 min
The best AI video generator in 2026 — an honest map Which AI video generator is best in 2026 depends on what you are making. An honest map by use case: cinematic clips with sound, motion, editing, free options, open source and same-character stories, with how each one really charges.
ai video & images · 2 min
Text to anime video — AI clips in an anime style, without a plan How to make anime-style AI video from a text description in 2026 — compose the frame first, animate it with sound, pay per clip instead of per month.
ai video & images · 2 min
AI music video — cinematic visuals for a track you already have How to make an AI music video in 2026: generate cinematic clips shot by shot, cut them to your track, pay per clip. The honest workflow, no subscription.
ai video & images · 2 min
Pika vs pay-as-you-go — when the monthly credit bucket stops making sense Pika's credits reset monthly and commercial rights start at the Pro plan. If you generate occasionally, pay-per-clip is cheaper — the honest math for 2026.
ai video & images · 2 min
Can Grok make videos? Yes — fast clips, daily quotas, and the alternative Yes. Grok makes video through Grok Imagine, xAI's own model — fast, stylized, and since early 2026 paid-only, behind X Premium or SuperGrok with daily quotas that reset. What it really costs, and the pay-per-clip alternative.
ai video & images · 2 min
Luma Dream Machine vs pay-as-you-go — the sound and the subscription Luma's plans start around US$30 a month, credits are opaque, and clips render silent. The pay-per-clip alternative with native sound — the honest math for 2026.
ai video & images · 4 min
Seedance 2.5 launched — but can you actually use it? Seedance 2.5 shipped July 31 2026 inside ByteDance's own apps, API still pending. What access really looks like from North America or Europe — and the pay-per-clip alternative that is open today.
ai video & images · 4 min
Can AI make a 10-minute video? The honest answer, 2026 No AI model renders ten coherent minutes in one go. How long a clip really runs in 2026, how extending and cast scenes stretch it, how many shots ten minutes takes, and what that costs, worked out honestly.
ai video & images · 2 min
Kling vs pay-as-you-go — when monthly credits stop making sense Kling sells monthly plans with credits that reset. If you generate occasionally, pay-per-clip is cheaper — the honest math, and where each approach wins in 2026.
ai video & images · 4 min
Text to video with sound — how native AI audio actually works Text-to-video AI that generates the sound with the picture, not silent clips with music added later. What native audio can do, what it still gets wrong, how to prompt for it, and a real clip to hear.
ai video & images · 3 min
How to use Veo 3 and Veo 3.1 without a Google account or Gemini subscription Veo 3 and Veo 3.1 need a Google account and a paid plan — and since May 2026 the Gemini app runs Omni instead. Clip lengths, what 3.1 added, and Veo-class video pay-as-you-go.
ai video & images · 2 min
AI video credits, explained — what a clip really costs What an AI video credit buys, what one clip costs in dollars, why daily-credit systems burn money, and the expiry fine print to check before buying a pack.
ai video & images · 3 min
Sora has shut down — here's what to do before September 24 OpenAI discontinued the Sora app on April 26, 2026, and the API died September 24. What happened, how to export your videos, and where pay-per-clip generation lives now.
ai video & images · 2 min
Cinematic AI video prompts — 12 ideas that actually render well What makes an AI video prompt cinematic: the anatomy that works, 12 scene fragments to steal, and why composing the frame first beats prompt engineering.
ai video & images · 2 min
How long does AI video generation take? Real numbers, 2026 Real AI video render times in 2026: what HD vs 4K actually takes, why queues on free tiers stretch to hours, and what the wait looks like clip by clip.
ai video & images · 2 min
Runway vs pay-as-you-go — which AI video pricing fits you What Runway credits cost per clip, what happens when they reset, and the pay-per-video math — when the monthly plan wins and when per-clip is cheaper.
ai video & images · 4 min
AI video generator with no subscription — how it works in 2026 What AI video costs without a monthly plan: how subscriptions, daily credits and free tiers really charge, the break-even maths for occasional creators, and what one clip costs pay-per-clip — with the first one free.

privacy & ephemeral ai

what ai tools keep, what they do not, and how to think out loud without leaving a trail.

privacy & ephemeral ai · 3 min
Is AI safe to use? Mainstream AI chatbots are safe to use for everyday tasks. The real risks are privacy, wrong answers, scams and over-reliance. What each one looks like in practice, which assistants keep what, and five habits that make any AI safer.
privacy & ephemeral ai · 4 min
ChatGPT data leaks — what actually happened, and what it means for you The real ChatGPT data leaks, plainly: the 2023 bug that showed users other people's chat titles and some payment details, the 2025 shared chats indexed by Google, the 2025 Mixpanel vendor breach, and the leak that happens most often: what people paste in.
privacy & ephemeral ai · 4 min
How to delete your ChatGPT history — every part of it How to delete one ChatGPT chat, every chat, archived chats, saved memories, shared links and your whole account, step by step, plus what deletion really does: the 30-day window, what survives, and why deleting a chat does not delete its memory.
privacy & ephemeral ai · 3 min
Is Gemini safe to use? Google's Gemini app is safe to install, but it keeps chats for 18 months by default, a subset is read by human reviewers, and reviewed chats can be kept for three years. The real risks, the settings that matter, and the Gemini crypto mix-up.
privacy & ephemeral ai · 3 min
Does ChatGPT sell your data? OpenAI says it does not sell personal data or share your conversations with advertisers. Since ads arrived in 2026, its privacy policy also describes ad personalisation and sharing with marketing partners. What that means, plainly, and how to opt out.
privacy & ephemeral ai · 3 min
ChatGPT privacy settings — every control that matters Every ChatGPT privacy setting worth changing in 2026: model training, temporary chat, memory and chat-history references, shared links, archived chats, deletion, data export, ad personalisation and connected apps. Where each one is, and what it actually does.
privacy & ephemeral ai · 4 min
AI privacy concerns — the real risks, and what to do about each The AI privacy concerns that actually matter in 2026: training on your chats, long retention, human review, court orders, ads, public share links, work accounts and where data is stored. Each with real examples and the setting or habit that addresses it.
privacy & ephemeral ai · 7 min
Signal vs thread — we measured it A pre-registered test of two ways an AI chat can keep context: the full transcript every turn, or a fixed intent plus a running log. 30 conversations, 708 turns, a blind AI judge. Where the log won, where the transcript won, and what it costs.
privacy & ephemeral ai · 3 min
Is Character.AI private? What happens to your chats Character.AI chats are private from other users, but they are stored, can be used to train its models and tailor ads, and can be read by staff for safety and legal reasons. What its own privacy policy says, and how to use it more carefully.
privacy & ephemeral ai · 3 min
Is Meta AI private? What Meta does with your chats Since December 16, 2025, Meta uses your conversations with Meta AI to personalise ads and content across Facebook, Instagram, WhatsApp and Messenger, with no full opt-out. What that covers, what is excluded, and what the Discover feed taught people.
privacy & ephemeral ai · 3 min
Is Perplexity private? What it keeps, and the setting that decides training Perplexity keeps your threads until you delete them and, on Free, Pro and Max, uses your searches to train its models unless you turn off AI data retention. What Incognito changes, what the 2026 lawsuit alleges, and how Enterprise differs.
privacy & ephemeral ai · 4 min
Is Grok private? What SpaceXAI keeps, who can read it, and what X changes Grok keeps your chats until you delete them, can train on them unless you switch it off, and some are reviewed by staff. What Private Chat changes, why share links once turned up in Google, and why Grok on X follows different rules.
privacy & ephemeral ai · 4 min
Is ChatGPT safe to use? The four risks that actually matter ChatGPT is not malware and the official app is safe to install. The real risks are quieter: confident wrong answers, what you paste into it, fake ChatGPT apps and sites, and what you let it connect to. How to use it safely, and for teens.
privacy & ephemeral ai · 5 min
Is ChatGPT private? What OpenAI keeps, and what it does with it ChatGPT keeps your chats until you delete them, can train on them unless you switch it off, and on free plans can use them to pick the ads you see. What each setting really controls, what temporary chat changes, and what deleting does.
privacy & ephemeral ai · 4 min
Is DeepSeek safe? It depends which DeepSeek you mean DeepSeek's own privacy policy says your chats are stored in China, kept for as long as you have an account, and used to train its models. What that means, why governments restricted it, and how running the open model yourself changes the answer.
privacy & ephemeral ai · 4 min
Is Gemini private? What Google keeps, who reads it, and for how long Gemini saves your chats for 18 months by default, some are read by human reviewers, and those reviewed chats are kept for up to three years even if you delete your activity. What Keep Activity really controls, what temporary chats change, and the 72 hours that never go away.
privacy & ephemeral ai · 5 min
Is Microsoft Copilot private? Two apps, two sets of rules Since August 18, 2026 there are two consumer Copilot apps with two sets of privacy rules. What each one keeps, where your chats feed personalised ads and Bing, what the old app said about training and human review, and why work accounts are different again.
privacy & ephemeral ai · 5 min
What can ChatGPT Enterprise and Business admins actually see? On a company ChatGPT workspace, your conversations are company records. What Enterprise and Business admins can actually access, what the Compliance API carries, why temporary chats are not a hiding place, and what the usage dashboard does not show.
privacy & ephemeral ai · 4 min
Is Claude private? What Anthropic keeps, and for how long Claude keeps your chats in your account until you delete them, and one privacy setting decides whether Anthropic can train on them and keep a de-identified copy for up to five years. What each choice means, what incognito changes, and what deleting actually does.
privacy & ephemeral ai · 6 min
What each AI chatbot keeps One table: what ChatGPT, Claude, Gemini, Copilot, Grok, DeepSeek, Duck.ai, Lumo, Brave Leo and nixm each retain, for how long, whether deleted means deleted, and what the training default is. Sourced to each vendor, dated, and updated monthly.
privacy & ephemeral ai · 5 min
Which AI chatbot is the most private? Anonymous, encrypted, no-retention, or local — five AI chat tools each win on a different definition of private. Lumo, Duck.ai, Brave Leo, Venice, running your own model, and the one that never stores a transcript at all.
privacy & ephemeral ai · 3 min
Does ChatGPT keep the files and images you upload? Photos, screenshots and files are stored on a separate clock from the chat. Delete the conversation and the PDF, screenshot or photo can still be there — and has to be removed on its own. What that means, and the version with no second list.
privacy & ephemeral ai · 3 min
ChatGPT record mode: where does the recording go? Record mode turns audio into a transcript and a canvas that stays in your chat history — and that ChatGPT can reference in later conversations. What OpenAI documents, and what it leaves you to arrange yourself.
privacy & ephemeral ai · 3 min
The case for a minimal AI interface Every setting an AI tool offers is a preference it has to store about you. The argument for interfaces that ask less — and keep less.
privacy & ephemeral ai · 3 min
How long does ChatGPT keep your chat history? About 30 days is the policy. A 2025 court order suspended that clock for months, and 20 million conversations were still handed to plaintiffs' lawyers. The real timeline, and an AI with no window at all.
privacy & ephemeral ai · 2 min
What data does ChatGPT store about you? Not just chats — account details, uploads, memory, device and usage signals. What actually accumulates behind an AI account, and what an AI that accumulates nothing looks like.
privacy & ephemeral ai · 2 min
Does ChatGPT record your conversations? People asking whether ChatGPT records them usually mean something specific: is someone listening, is it training on me, can this be pulled up later. Three different answers.
privacy & ephemeral ai · 2 min
An AI with no memory — feature, not bug Every AI company is racing to remember you harder. The case for the opposite: what memory actually costs, what forgetting buys, and when a blank-slate AI is the better tool.
privacy & ephemeral ai · 2 min
Is your journaling app analyzing you? Mood graphs, theme detection, pattern tracking — the features that make AI journals feel smart all require reading everything you wrote. What analysis means for your archive.
privacy & ephemeral ai · 2 min
The case for ephemeral journaling Journaling advice assumes you keep the journal. A serious case for the opposite: writing that dissolves, why it produces more honesty, and what you actually lose.
privacy & ephemeral ai · 2 min
The 5-minute brain-dump that dissolves behind you A simple evening routine: dump everything circling in your head into an AI that sorts it with you — then let the whole session dissolve. Five minutes, no transcript kept, repeatable forever.
privacy & ephemeral ai · 2 min
Private journaling apps — what each model actually stores Every journaling app claims privacy — encrypted, on-device, passcoded. A plain map of what each model actually stores and where, and the one option that keeps no transcript at all.
privacy & ephemeral ai · 2 min
A private space to think out loud — that keeps nothing Everyone needs somewhere to think unedited. Why AI is good at holding one end of that, why stored history ruins it, and what a room that keeps no transcript changes.
privacy & ephemeral ai · 2 min
Venting to an AI — here's what happens to the words Millions vent to AI after hard days. On most tools, the vent becomes stored history under your account. What actually happens to those words — and where they can happen without a record.
privacy & ephemeral ai · 2 min
Prompts for a tough day — in a room that saves nothing Seven prompts for writing out a hard day — designed for a room that keeps no transcript, where the honest answer is safe to write because no lasting record of it will exist.
privacy & ephemeral ai · 2 min
AI thinking partner vs. therapist — the line, drawn honestly AI can be a genuinely useful thinking partner — and it is not therapy. The honest line between the two, why the distinction protects you, and when only a human will do.
privacy & ephemeral ai · 2 min
AI journaling without the data trail Popular AI journals are memory-first — entries stored, indexed, analyzed over months. The other model: journal with an AI that keeps no transcript at all, trace dissolving in 7 days.
privacy & ephemeral ai · 2 min
Is ChatGPT confidential for work? Feels yes, is no It feels like a private conversation; it's a stored business record. The five confidentiality mistakes professionals make with AI, and the two-lane fix that ends them.
privacy & ephemeral ai · 2 min
Can your employer see your ChatGPT? Depends which ChatGPT Work account, personal account, work laptop, work network — each changes who can see your AI chats. The visibility map, honestly drawn, and the option outside all of it.
privacy & ephemeral ai · 2 min
Sensitive decisions — brainstormed where nothing sticks The hardest decisions need the messiest thinking — and the messiest thinking least deserves a permanent record. A room for working out what you actually think, that keeps none of it.
privacy & ephemeral ai · 2 min
Negotiation prep — where your walk-away number shouldn't live Your walk-away number, your fallback positions, your read of the other side — negotiation prep is leverage, and stored AI history is where leverage goes to leak. Prep off the record.
privacy & ephemeral ai · 2 min
M&A thinking with AI — what not to paste, and where to think Deal thinking leaks through tooling before it leaks through people. What M&A material should never enter stored AI, what generalizes safely, and where the modeling can live.
privacy & ephemeral ai · 2 min
HR drafts and AI — where the most sensitive text in the company goes PIPs, terminations, investigation notes — HR work is the most sensitive text in a company, and AI drafting files it to consumer chat history by default. Where it goes, and where it doesn't have to.
privacy & ephemeral ai · 2 min
Layoff planning with AI — without the thinking becoming a record Layoff planning is the most leak-sensitive document work a leader does — and consumer AI stores every draft. How to think through a reduction without creating a record of the thinking.
privacy & ephemeral ai · 2 min
AI and privilege — the reality nobody's marketing department mentions AI chats carry no privilege — Altman said it plainly, the ABA warned about it, and courts proved it. What lawyers and clients get wrong, and what reducing the trail actually means.
privacy & ephemeral ai · 2 min
What people delete from ChatGPT — and the better move than deleting Deleting AI chats after the fact is the most common privacy habit — and the weakest. What the delete-later reflex reveals, why it fails, and the ask-without-a-record alternative.
privacy & ephemeral ai · 4 min
Private AI chat — no account, and as little kept as possible What private AI chat can actually mean in 2026: no account, no stored history, no training, or encryption. Which tools offer which, why no-account alone is not enough, and what nixm keeps, for how long, stated plainly.
privacy & ephemeral ai · 3 min
Ephemeral AI — when the conversation never becomes a record Ephemeral AI chat means no transcript is kept, and whatever the service does keep is small and ends on a fixed clock. How it differs from no-logs and incognito modes, what nixm keeps for seven days, and why the design exists.
privacy & ephemeral ai · 4 min
NYT v. OpenAI — how a copyright case became about your chats A copyright lawsuit turned into the biggest AI-privacy story of the decade: deleted chats preserved, 20 million conversations produced, and a July 2026 sanctions motion alleging logs were deleted anyway.
privacy & ephemeral ai · 2 min
Relationship questions — thought through, not filed away People ask AI about their relationships constantly — and every question becomes account history. How to think through the personal stuff without a permanent record.
privacy & ephemeral ai · 2 min
The questions you'd never Google — asked at last Search history trained everyone to self-censor. AI history is deeper — full sentences, real context. Where the never-Google questions can finally be asked: a room that keeps no transcript.
privacy & ephemeral ai · 2 min
No-log AI — how to read the claim like a skeptic Every private AI says "no logs" — and each means something different. How to read the claims: what's encrypted, what's local, what's anonymized, what's truly never stored.
privacy & ephemeral ai · 2 min
Incognito isn't enough — where AI privacy actually lives Incognito thinking — hide it from my screen — misses where AI privacy actually lives: the server. The three layers real privacy requires, and where each tool stops.
privacy & ephemeral ai · 2 min
How to ask AI personal questions — privately, actually A practical, layered playbook for asking AI the personal stuff — what settings help, what they miss, what not to include, and the option where nothing persists at all.
privacy & ephemeral ai · 3 min
Does ChatGPT save my conversations? Yes — here's the whole picture Yes — ChatGPT saves your conversations, and a federal court spent a year deciding what happens to them. What's retained, for how long, and what an AI that keeps no transcript looks like.
privacy & ephemeral ai · 3 min
Encrypted vs. never stored — protecting the record vs. having none Zero-access encryption protects a record that exists; ephemerality means no transcript is kept and what little remains dissolves on a clock. Why the distinction matters for breaches, subpoenas, and time — explained honestly.
privacy & ephemeral ai · 2 min
Duck.ai — genuinely anonymous, differently private Duck.ai anonymizes who you are, not what persists. What DuckDuckGo's own docs say it strips and keeps, where your chats actually go, and what anonymity doesn't cover.
privacy & ephemeral ai · 2 min
The awkward questions — asked without a record The questions people won't type into a search bar are the ones most worth asking. What gets stored when you ask them, which settings actually stop it, and what a chat that keeps no transcript looks like.
privacy & ephemeral ai · 2 min
How to use AI without it remembering you Practical steps to keep AI from accumulating a record of you — settings that help, their limits, and the one option where forgetting is the architecture.
privacy & ephemeral ai · 2 min
Can AI companies read your chats? Where the access really sits Who can actually access your AI conversations — company systems, reviewers, courts, breaches — and what changes when there is no stored conversation to access.
privacy & ephemeral ai · 3 min
Can deleted AI chats end up in court? They already have Courts treat AI conversations like email — preservable, discoverable, producible by the millions. What the OpenAI litigation established, why deletion isn't protection, and what has nothing to produce.
privacy & ephemeral ai · 2 min
ChatGPT Temporary Chats — private from you, not from the server Temporary Chats hide conversations from your history — not from OpenAI's servers. The 30-day window, the court order that swept them up, and what actual ephemerality means.
privacy & ephemeral ai · 2 min
AI chat that doesn't save history — an honest map of the field Lumo, Venice, Duck.ai, Brave Leo, temporary modes — each "private AI" stores something different. An honest map of what each keeps, and the one architecture that keeps no transcript.

gifting

gifting · 3 min
Can you gift Claude? How Claude gift subscriptions work Yes, you can gift Claude Pro or Max for 1, 3, 6 or 12 months, by email or a shareable link. The catch: only paying Pro and Max subscribers can buy one. How it works, what the recipient gets, the 365-day expiry, and what to do if you are not a subscriber.
gifting · 3 min
Can you gift Gemini? Google does not sell a Gemini gift card or gift subscription. The real way to give someone Google AI Pro is to add them to your Google family group: up to five people, each with their own Gemini limits. How it works, the rules, and the alternatives.
gifting · 3 min
AI gift ideas — what you can actually give Practical AI gift ideas for 2026: a ChatGPT gift card, a Claude gift subscription, an AI video gift, or a short film made of someone you love. Which ones anyone can buy, which need a subscription first, and what Google does not sell.
gifting · 5 min
Is there an OpenAI gift card? As of September 2026 OpenAI sells ChatGPT gift cards through US retailers like Best Buy. They never expire — unlike gift credits, which do. What the balance can and cannot buy, who can redeem one, and the alternative that works outside the US.
gifting · 4 min
How to gift ChatGPT credits, step by step The full flow from OpenAI's own documentation: who can buy, how the claim link works, the 30-day and 12-month clocks — and when the new US-only gift card (no expiry) is the better choice instead.
gifting · 4 min
Do gifted ChatGPT credits expire? Redeemed credits expire 12 months after redemption and an unclaimed link dies at 30 days. OpenAI's newer gift card does not expire at all — but is US-only. What that means if you are buying in advance.
gifting · 4 min
Can you gift a ChatGPT subscription? You cannot gift the plan itself, but OpenAI's US-only gift card ($15–$250, no expiry) adds a balance that pays for Plus or Pro on the web. Gift credits expire 12 months after redemption. Which to buy, and the option that works outside the US — CA$24.99.

notes

arguments, not guides — on intent, craft, and what these tools are actually for.

notes · 4 min
Six weeks of solo-founder SEO — the real numbers A first-hand report from building nixm alone: which directories were worth paying for, what Domain Rating really did, why privacy pages indexed in a day while video pages sat uncrawled for weeks, the impressions curve, and what got our accounts flagged.
notes · 4 min
ai video is a slot machine — until you decide something The slot-machine criticism of AI video is right about one-line prompts and wrong about the rest. On intent, delegation, and where slop really comes from.

Enter your email and receive a link to sign in. No passwords, just a secure signal to connect.

why connect

20 free credits to start
your signals, remembered
buy more credits anytime
image & video generation features

three steps. one nixm.

type it. it locks. you share it.

no setup. no onboarding. no prompt engineering.

step 01:
seed your intent

open nixm^™. type what's actually on your mind — not a command, just your intent. raw and direct. you're setting the frequency.

"prep for jack's party this weekend"
"think through our brand direction for q1"
"build me a qr generator"

step 02:
get your nixm

nixm^™ locks to your intent and hands you back a nixm — a URL that holds your primed intelligence. anyone who opens it steps into the same context, immediately.

NIXM-JACK-P4RT NIXM-ATR1-GAME NIXM-SURR-DRMS

step 03:
share or return

send it to your team — they step into the same frequency instantly, no setup. or keep it private: the nixm holds your context until you return.

→ share it with anyone
→ or save it for yourself
→ the nixm doesn't care — it just holds

start one

the fastest way to understand a nixm is to seed one. close this and type what's actually on your mind — that's the whole first step. seed a nixm →

trois étapes. un nixm.

vous l'écrivez. il se verrouille. vous le partagez.

aucune configuration. aucun tutoriel. aucune ingénierie de prompt.

étape 01 :
amorcez votre intention

ouvrez nixm^™. écrivez ce que vous avez vraiment en tête — pas une commande, juste votre intention. brute et directe. c'est vous qui réglez la fréquence.

« préparer la fête de jack ce week-end »
« réfléchir à notre direction de marque pour le t1 »
« construis-moi un générateur de qr »

étape 02 :
recevez votre nixm

nixm^™ se verrouille sur votre intention et vous rend un nixm — une URL qui porte votre intelligence déjà amorcée. quiconque l'ouvre entre dans le même contexte, immédiatement.

NIXM-JACK-P4RT NIXM-ATR1-GAME NIXM-SURR-DRMS

étape 03 :
partagez ou revenez

envoyez-le à votre équipe — elle entre instantanément sur la même fréquence, sans configuration. ou gardez-le privé : le nixm conserve votre contexte jusqu'à votre retour.

→ partagez-le avec qui vous voulez
→ ou gardez-le pour vous
→ le nixm s'en moque — il retient, c'est tout

amorcez-en un

le moyen le plus rapide de comprendre un nixm, c'est d'en amorcer un. fermez ceci et écrivez ce que vous avez vraiment en tête — c'est toute la première étape. amorcer un nixm →

common questions.

what exactly is a nixm?

a nixm is a unique URL that carries your primed AI context with it. you seed it with intent, it locks to that frequency, and anyone who opens the link enters the same focused intelligence you created. no setup, no re-explaining.

nixm^™, a nixm, a signal — which is which?

three words, three things.

nixm^™ is the interface — the thing you're using right now.

a nixm is what you make with it: a URL that holds your intent. seed one, share it, come back to it.

a signal is the intent itself, locked in place. every nixm carries one — that's what makes it hold. (under the hood it's primed on two: an intent signal and a collab signal. that's the machinery, not something you set.)

short version: you use nixm^™ to make a nixm, and the signal is what's locked inside it.

can nixm show me things — charts, visuals, demos?

yes. ask nixm^™ to show, visualize, render, or animate something and it renders live inline — particles, 3D scenes, charts, interactive demos. data and ideas, made visible. for code you want to keep, copy, or modify, ask it to build something instead — that gives you the full source with a toolbar to expand, copy, and export. for pictures, ask it to generate an image — square, landscape, or portrait, rendered inline.

can i animate a still?

any still in a signal can be set in motion — say "animate this" (or press ANIMATE on the image) and nixm^™ builds a motion pass from that exact frame: one camera move, ambient action, sound if the scene implies it. the frame stays the frame. one video credit.

can i make a video longer?

yes. every finished clip has an EXTEND button — about seven more seconds, same unbroken shot, one video credit per pass. even better: ask in words. "keep going, but have her turn toward the light" gives the extension a direction, and a directed continuation almost always beats a neutral one.

what is CAST?

a signal can hold a cast — characters you've established with reference images. upload a face, or use a frame nixm^™ already rendered, give it a name, and it stays. then put that character in a new scene: a different place, a different action, same face. nixm calls it a scene; you'll see the shot and the price before anything is spent.

how do i add someone to the cast?

attach an image in the conversation and tell nixm^™ who it is. one image at a time — nixm asks whether each new one is a new character or another angle of someone you already have. a character can hold up to four images (one main, three angles), which is what keeps the face consistent.

what can i upload?

images you own or have the right to use — photos of yourself, people who have agreed to be there, things you shot, frames nixm^™ already rendered for you. one at a time, under 12mb.

what does not belong here: anyone under 18, anything sexual, real people who have not agreed — public figures included — and characters or footage someone else owns. a costumed superhero is somebody's property even when a model draws it happily. the model's filter is not the rule; this is.

uploads are screened automatically and spot-checked by a person. anything outside those lines is removed, and the account with it. everything else dissolves with the signal, as usual.

what does a scene cost?

10 seconds is 2 video credits, 15 seconds is 3, however many characters are in it. sound is included in both. nixm^™ shows you the cast, the shot and the price first — nothing renders until you say go.

how long does a scene take?

about 10 to 15 minutes, whichever length you pick — these are bigger renders than a standard clip. you can close the page; the clip lands on the signal when it's done.

how many characters can a scene hold?

a scene holds four references — cast and props together. three characters and a prop, two characters and two props, four faces. more than that and someone doesn't make it into the frame; nixm^™ proposes inside the limit and says what it cut. each character can carry up to three angles — one photo works, three hold better.

what can a scene actually hold?

less than a sentence can ask for. two camera positions in fifteen seconds, one in ten — a third shot costs a face. a character arrives in the clothes from their reference; text won't change them. to dress someone, cast the outfit as a reference and say wearing it, or render a still of them in the new clothes and cast that. one dressed character per scene is what holds today, and a print on a shirt needs a medium shot or closer to read. nixm knows these limits and proposes within them — when your ask runs past, it tells you what it trimmed, or offers two scenes.

can a character talk?

yes, and the voice lands — but lip sync is still experimental, even on a still face looking at the lens. a moving mouth in a moving shot won't match. if the line matters, keep the shot still; if the motion matters, let the sound carry it. this improves with the engine; we'll say when it does.

can i cast a face nixm^™ rendered, not one i uploaded?

yes. render the still, then say cast her as vesper — naming it is what makes it a cast member. it moves into the cast strip and works exactly like an upload.

can i rename a character?

yes, any time — just ask. the new name sticks to that character and nixm^™ uses it from then on.

do my uploaded references stay private?

they live in the signal, like everything else you make there — and they dissolve with it. a public signal shows its cast to anyone with the link; mark the signal private and it's yours alone. either way, nothing is trained on and nothing outlives the signal. one exception, stated plainly: uploads are screened for safety, and a person may look at an image for that reason — never at your dialogue.

what does "seeding" mean?

seeding is how you set the frequency. instead of just asking a question, you give nixm^™ context: your goal, your mindset, your constraints. it locks to that — and every response from then on stays anchored to what you actually wanted.

how is this different from a chatbot?

a chatbot is stateless — it treats every message as a fresh transaction. nixm^™ uses a contextual anchor that keeps your original intent alive across the whole session. it doesn't drift. it doesn't forget what you came to do.

why does the interface change color? can i pick my own?

the color shift is a live readout of the session's state — thinking, idle, reacting. it's a small feedback loop between the AI's processing and your visual field. signal activity made visible. want a different color? just ask — "use green", "tone down the background".

what happens when a signal expires?

signals live for seven days from creation. after that, the signal dissolves — its intent, its distilled log and any images you uploaded — and nothing of it remains. nixm^™ is built on the principle of ephemeral intelligence: no transcript of your conversation is ever kept. the insights are yours. the noise is deleted. nothing is harvested.

can a signal live longer than seven days?

yes — now it can. one credit adds seven days to a signal you own. press "+7 days" in your dash as many times as you like; the days stack. dissolving stays the default — keeping something alive is a choice you make, not a setting you forget.

can i end a signal before seven days?

yes. press "dissolve" in your dash, beside the privacy toggle — it appears only on signals you own. one tap arms it, a second confirms, and the dialogue, its images and its clips go immediately. private hides a signal; dissolve ends it. there is no undo, which is the point.

made a signal without signing in and want it gone early?

signals made without an account have no owner, so there's no dissolve button. email info@nixm.ai with the signal's id (NIXM-…) and we'll dissolve it by hand — nothing else about you is needed. do nothing and it dissolves on its own seven days after it was made.

what survives after a signal dissolves?

whatever you took out. the frames and clips you downloaded are yours forever — the conversation is the part that expires. nothing is kept, nothing is harvested. that's not a limitation of the product; it is the product.

is nixm^™ open source?

no. the orchestration layer, contextual anchor logic, and session architecture are proprietary. this isn't a wrapper — it's a custom-built framework designed to keep your intent alive across a session in a way standard AI cannot.

is nixm^™ private?

structurally, yes. nixm^™ doesn't keep a transcript of your conversation. a distilled log stays with the signal for seven days, then dissolves — and nothing you type is used for training. and if you need something from a past session — just ask again. your thinking stays yours because we never kept a transcript of it.

one clarification: a shared nixm carries your intent seed publicly — that's the feature. anyone with the link enters the same focused context. what stays private is the dialogue itself: no transcript of it is kept. the signal's distilled log is part of that shared context, and it dissolves with the signal after seven days. you control what you seed. nixm^™ controls nothing after that.

can i share a nixm with someone else?

yes. every nixm is shareable by default. when you create a nixm, you can share the link and anyone who opens it enters the same focused intelligence you created. they don't need to set up context — the nixm holds it.

are my signals public or private?

by default, signals you create while logged out are public — anyone with the link can open them. once you log in, you can mark any signal you own as private: the link still works only for you, never for anyone else. toggle privacy from the signal ID pill at the top of the screen when you're signed in.

what does everything cost?

a message costs 1 credit. an image adds 2 more — charged only when the image actually arrives. video runs on its own separate pool: 1 video credit for HD, 2 for 4K, charged when the clip is delivered. a cinematic scene with your cast is 2 credits at 10 seconds, 3 at 15. extending a clip is 1 video credit per pass; keeping a signal alive is 1 credit per +7 days.

what if a render fails?

you pay for the message, never for media that didn't arrive. a failed or filtered render doesn't charge the image or video part — that's charged at delivery, on purpose.

do credits expire?

no. never. they'll be waiting whenever the next idea is.

how do i redeem a gift code?

sign in (free — a magic link, no password), open the sign-in panel, and tap have a gift code?. enter the code — credits land instantly on your account. gift codes never expire, and one code redeems once, on any account.

i lost my gift code or the email — can it be recovered?

yes. the code was emailed to the buyer (and to the recipient, if one was added at checkout) — search your inbox for "nixm". if the email is gone, the buyer can write to info@nixm.ai with the purchase email address and we'll look it up. unredeemed codes stay valid indefinitely.

are downloaded files safe to run?

nixm^™ produces real, executable code. before running anything you download, open it in an editor and read it first. treat it like code from any source on the internet — not magic, not malware, just code. use your discretion.

what can i actually ask nixm to do?

more than people expect. nixm^™ isn't a chatbot — it's a primed surface for shaping work. some examples to try:

what	try this
build a tool	build a beatbox in vanilla js. build a QR code generator. build a markdown editor with live preview.
visualize an idea	visualize photosynthesis in 3d. show me q1 growth as a chart. animate how a transformer works.
write a brief or doc	write a brand brief in editorial tone. draft a one-page hiring rubric. summarize this thread into a memo.
think out loud	walk me through pricing for a 50-person agency. compare two technical approaches. roast my product copy.
study or learn	explain mitochondria like i'm 14. quiz me on the krebs cycle. visualize the periodic table by reactivity.
iterate	refine the last response, but tighter. give me a more cynical version. cut everything that's not specific.

if it can be described, you can ask it. nixm^™ answers the way the signal is primed.

questions fréquentes.

qu'est-ce qu'un nixm, exactement ?

un nixm est une URL unique qui transporte avec elle votre contexte d'IA déjà amorcé. vous l'amorcez avec une intention, il se verrouille sur cette fréquence, et quiconque ouvre le lien entre dans la même intelligence focalisée que vous avez créée. aucune configuration, aucune explication à refaire.

nixm^™, un nixm, un signal — lequel est lequel ?

trois mots, trois choses.

nixm^™ est l'interface — ce que vous utilisez en ce moment même.

un nixm est ce que vous créez avec : une URL qui porte votre intention. amorcez-en un, partagez-le, revenez-y.

un signal est l'intention elle-même, verrouillée. chaque nixm en porte un — c'est ce qui le fait tenir. (en coulisses, il est amorcé sur deux : un signal d'intention et un signal de collab. c'est la mécanique, rien que vous ayez à régler.)

en bref : vous utilisez nixm^™ pour créer un nixm, et le signal est ce qui est verrouillé dedans.

est-ce que nixm peut me montrer des choses — graphiques, visuels, démos ?

oui. demandez à nixm^™ de montrer, visualiser, rendre ou animer quelque chose et le résultat s'affiche en direct dans le fil — particules, scènes 3D, graphiques, démos interactives. les données et les idées, rendues visibles. pour du code que vous voulez conserver, copier ou modifier, demandez-lui plutôt de construire quelque chose — vous obtenez alors la source complète, avec une barre d'outils pour agrandir, copier et exporter. pour des images, demandez-lui de générer une image — carrée, paysage ou portrait, affichée dans le fil.

puis-je animer une image fixe ?

n'importe quelle image fixe d'un signal peut être mise en mouvement — dites « anime ça » (ou appuyez sur ANIMATE sur l'image) et nixm^™ construit une passe de mouvement à partir de cette image exacte : un mouvement de caméra, une action d'ambiance, du son si la scène l'appelle. l'image reste l'image. un crédit vidéo.

puis-je allonger une vidéo ?

oui. chaque clip terminé porte un bouton EXTEND — environ sept secondes de plus, le même plan ininterrompu, un crédit vidéo par passe. mieux encore : demandez-le avec des mots. « continue, mais fais-la se tourner vers la lumière » donne une direction à l'extension, et une continuation dirigée vaut presque toujours mieux qu'une continuation neutre.

qu'est-ce que CAST ?

un signal peut porter une distribution — des personnages que vous avez établis à partir d'images de référence. téléversez un visage, ou reprenez une image que nixm^™ a déjà rendue, donnez-lui un nom, et il reste. vous pouvez ensuite placer ce personnage dans une nouvelle scène : un autre lieu, une autre action, le même visage. nixm appelle ça une scène ; vous verrez le plan et le prix avant que quoi que ce soit ne soit dépensé.

comment ajouter quelqu'un à la distribution ?

attachez une image dans la conversation et dites à nixm^™ de qui il s'agit. une image à la fois — nixm demande si chaque nouvelle image est un nouveau personnage ou un autre angle de quelqu'un que vous avez déjà. un personnage peut porter jusqu'à quatre images (une principale, trois angles), et c'est ce qui garde le visage constant.

que puis-je téléverser ?

des images que vous possédez ou que vous avez le droit d'utiliser — vos propres photos, des personnes qui ont donné leur accord, ce que vous avez filmé ou photographié, des images que nixm^™ a déjà produites pour vous. une à la fois, sous 12 mo.

ce qui n'a pas sa place ici : toute personne de moins de 18 ans, tout contenu sexuel, toute personne réelle qui n'a pas donné son accord — personnalités publiques comprises — et tout personnage ou extrait appartenant à autrui. un super-héros costumé appartient à quelqu'un, même quand un modèle le dessine sans broncher. le filtre du modèle n'est pas la règle ; celle-ci l'est.

les téléversements font l'objet d'un contrôle automatique et de vérifications humaines ponctuelles. tout ce qui sort de ces limites est retiré, et le compte avec. le reste se dissout avec le signal, comme d'habitude.

combien coûte une scène ?

10 secondes valent 2 crédits vidéo, 15 secondes en valent 3. le son est inclus dans les deux cas. nixm^™ vous montre d'abord la distribution, le plan et le prix — rien ne se rend tant que vous n'avez pas dit d'y aller.

combien de temps prend une scène ?

environ 10 à 15 minutes, quelle que soit la durée choisie — ce sont des rendus plus lourds qu'un clip standard. vous pouvez fermer la page ; le clip arrive sur le signal une fois terminé.

combien de personnages tient une scène ?

une scène tient quatre références — distribution et accessoires compris. trois personnages et un accessoire, deux et deux, quatre visages. au-delà, quelqu'un manque à l'image ; nixm^™ propose dans la limite et dit ce qu'il a retiré. chaque personnage peut porter jusqu'à trois angles — une photo suffit, trois tiennent mieux.

que peut vraiment tenir une scène ?

moins qu'une phrase peut demander. deux positions de caméra en quinze secondes, une en dix — un troisième plan coûte un visage. un personnage arrive dans les vêtements de sa référence ; le texte ne les change pas. pour l'habiller, ajoutez la tenue comme référence et dites il la porte, ou rendez une image de lui dans la nouvelle tenue et castez-la. un personnage habillé par scène, c'est ce qui tient aujourd'hui, et un imprimé sur un t-shirt se lit en plan moyen ou plus serré. nixm connaît ces limites et propose à l'intérieur — quand la demande déborde, il dit ce qu'il a coupé, ou propose deux scènes.

puis-je renommer un personnage ?

oui, à tout moment — il suffit de le demander. le nouveau nom reste attaché à ce personnage et nixm^™ l'utilise à partir de là.

mes références téléversées restent-elles privées ?

elles vivent dans le signal, comme tout ce que vous y créez — et elles se dissolvent avec lui. un signal public montre sa distribution à quiconque a le lien ; marquez le signal comme privé et elle n'appartient qu'à vous. dans les deux cas, rien ne sert à entraîner quoi que ce soit et rien ne survit au signal. une exception, dite clairement : les téléversements font l'objet d'un contrôle de sécurité, et une personne peut regarder une image à ce titre — jamais votre dialogue.

que veut dire « amorcer » ?

amorcer, c'est régler la fréquence. au lieu de simplement poser une question, vous donnez un contexte à nixm^™ : votre objectif, votre état d'esprit, vos contraintes. il se verrouille dessus — et chaque réponse reste ensuite ancrée à ce que vous vouliez vraiment.

en quoi est-ce différent d'un chatbot ?

un chatbot est sans état — il traite chaque message comme une transaction isolée. nixm^™ utilise un ancrage contextuel qui maintient votre intention d'origine vivante sur toute la session. il ne dérive pas. il n'oublie pas ce que vous êtes venu faire.

pourquoi l'interface change-t-elle de couleur ? puis-je choisir la mienne ?

le glissement de couleur est une lecture en direct de l'état de la session — réflexion, veille, réaction. c'est une petite boucle de retour entre le traitement de l'IA et votre champ visuel. l'activité du signal, rendue visible. vous en voulez une autre ? demandez-le simplement — « passe au vert », « adoucis le fond ».

que se passe-t-il quand un signal expire ?

les signaux vivent sept jours à compter de leur création. passé ce délai, le signal se dissout — son intention, son journal condensé et les images que vous avez importées — et il n'en reste rien. nixm^™ repose sur le principe d'une intelligence éphémère : aucune transcription de votre conversation n'est jamais conservée. les idées sont à vous. le bruit est supprimé. rien n'est exploité.

un signal peut-il vivre plus de sept jours ?

oui — c'est désormais possible. un crédit ajoute sept jours à un signal qui vous appartient. appuyez sur « +7 days » dans votre tableau de bord autant de fois que vous le souhaitez ; les jours s'additionnent. la dissolution reste le comportement par défaut — garder quelque chose en vie est un choix que vous faites, pas un réglage que vous oubliez.

puis-je mettre fin à un signal avant sept jours ?

oui. appuyez sur « dissoudre » dans votre tableau de bord, à côté du réglage de confidentialité — il n'apparaît que sur les signaux qui vous appartiennent. une pression l'arme, une seconde confirme, et le dialogue, ses images et ses clips disparaissent immédiatement. privé le masque ; dissoudre y met fin. il n'y a pas de retour en arrière, et c'est voulu.

signal créé sans compte : comment le supprimer plus tôt ?

un signal créé sans compte n'a pas de propriétaire, donc pas de bouton « dissoudre ». écrivez à info@nixm.ai avec l'identifiant du signal (NIXM-…) et nous le dissoudrons à la main — rien d'autre sur vous n'est nécessaire. sinon, il se dissout tout seul sept jours après sa création.

que reste-t-il après la dissolution d'un signal ?

ce que vous en avez sorti. les images et les clips que vous avez téléchargés sont à vous pour toujours — c'est la conversation qui expire. rien n'est conservé, rien n'est récolté. ce n'est pas une limite du produit ; c'est le produit.

nixm^™ est-il open source ?

non. la couche d'orchestration, la logique d'ancrage contextuel et l'architecture de session sont propriétaires. ce n'est pas un wrapper — c'est un cadre sur mesure conçu pour garder votre intention vivante sur toute une session, ce que l'IA standard ne sait pas faire.

nixm^™ est-il privé ?

structurellement, oui. nixm^™ ne conserve aucune transcription de votre conversation. un journal condensé reste avec le signal pendant sept jours, puis se dissout — et rien de ce que vous tapez ne sert à entraîner un modèle. et si vous avez besoin de quelque chose d'une session passée — redemandez, simplement. votre pensée reste la vôtre parce que nous n'en gardons jamais la transcription.

une précision : un nixm partagé transporte publiquement votre amorce d'intention — c'est justement la fonctionnalité. quiconque a le lien entre dans le même contexte focalisé. ce qui reste privé, c'est le dialogue lui-même : aucune transcription n'en est conservée. le journal condensé du signal fait partie de ce contexte partagé, et il se dissout avec le signal au bout de sept jours. vous contrôlez ce que vous amorcez. nixm^™ ne contrôle rien au-delà.

puis-je partager un nixm avec quelqu'un d'autre ?

oui. chaque nixm est partageable par défaut. quand vous créez un nixm, vous pouvez en partager le lien, et quiconque l'ouvre entre dans la même intelligence focalisée que vous avez créée. la personne n'a aucun contexte à mettre en place — le nixm le retient.

mes signaux sont-ils publics ou privés ?

par défaut, les signaux que vous créez sans être connecté sont publics — quiconque a le lien peut les ouvrir. une fois connecté, vous pouvez marquer comme privé n'importe quel signal qui vous appartient : le lien ne fonctionne alors que pour vous, jamais pour quelqu'un d'autre. la confidentialité se règle depuis la pastille d'identifiant du signal, en haut de l'écran, lorsque vous êtes connecté.

combien coûte l'ensemble ?

un message coûte 1 crédit. une image en ajoute 2 — facturés seulement quand l'image arrive vraiment. la vidéo fonctionne sur une réserve séparée : 1 crédit vidéo en HD, 2 en 4K, facturés à la livraison du clip. une scène cinématique avec votre distribution coûte 2 crédits à 10 secondes, 3 à 15. prolonger un clip coûte 1 crédit vidéo par passe ; garder un signal en vie coûte 1 crédit par tranche de +7 jours.

et si un rendu échoue ?

vous payez le message, jamais un média qui n'est pas arrivé. un rendu qui échoue ou qui est filtré ne facture pas la partie image ou vidéo — celle-ci est facturée à la livraison, délibérément.

les crédits expirent-ils ?

non. jamais. ils vous attendront jusqu'à la prochaine idée.

comment utiliser un code cadeau ?

connectez-vous (c'est gratuit — un lien magique, sans mot de passe), ouvrez le panneau de connexion et touchez un code cadeau ?. saisissez le code — les crédits arrivent immédiatement sur votre compte. les codes cadeaux n'expirent jamais, et un code s'utilise une seule fois, sur n'importe quel compte.

j'ai perdu mon code cadeau ou l'e-mail — peut-on le retrouver ?

oui. le code a été envoyé par e-mail à l'acheteur (et au destinataire, si l'un a été ajouté au moment du paiement) — cherchez « nixm » dans votre boîte de réception. si l'e-mail a disparu, l'acheteur peut écrire à info@nixm.ai avec l'adresse e-mail utilisée pour l'achat et nous le retrouverons. les codes non utilisés restent valables indéfiniment.

les fichiers téléchargés peuvent-ils être exécutés sans risque ?

nixm^™ produit du code réel, exécutable. avant de lancer quoi que ce soit que vous avez téléchargé, ouvrez-le dans un éditeur et lisez-le d'abord. traitez-le comme du code venant de n'importe quelle source sur internet — ni magie, ni logiciel malveillant, juste du code. à vous de juger.

que puis-je réellement demander à nixm ?

plus de choses qu'on ne l'imagine. nixm^™ n'est pas un chatbot — c'est une surface amorcée pour façonner du travail. quelques exemples à essayer :

quoi	essayez ceci
construire un outil	construis une boîte à rythmes en js pur. construis un générateur de QR code. construis un éditeur markdown avec aperçu en direct.
visualiser une idée	visualise la photosynthèse en 3d. montre-moi la croissance du t1 sous forme de graphique. anime le fonctionnement d'un transformeur.
rédiger un brief ou un document	rédige un brief de marque sur un ton éditorial. rédige une grille d'embauche d'une page. résume ce fil en une note.
réfléchir à voix haute	explique-moi la tarification pour une agence de 50 personnes. compare deux approches techniques. démolis le texte de mon produit.
étudier ou apprendre	explique-moi les mitochondries comme si j'avais 14 ans. interroge-moi sur le cycle de krebs. visualise le tableau périodique par réactivité.
itérer	affine la dernière réponse, mais plus serrée. donne-moi une version plus cynique. coupe tout ce qui n'est pas précis.

si ça peut se décrire, ça peut se demander. nixm^™ répond à la mesure de l'amorce du signal.

conditions of use.

nixm^™ is an experimental environment. by entering a signal, you acknowledge the transient nature of the data and accept these conditions.

intellectual property. the nixm^™ interface logic, contextual anchor framework, and session architecture are proprietary. unauthorized replication of this architecture is prohibited.

what you upload. you can attach an image to a signal — as a reference for a character, or as a frame to set in motion. by uploading, you confirm three things: you made it or you hold the rights to use it; every identifiable person in it is an adult who has agreed to appear; and it contains none of the material listed below. you keep ownership of what you upload. you grant us only what running the service requires — storing the file, passing it to the model providers that render your work, and displaying it inside your signal — for as long as that signal lives, and no longer. we do not train on it and we do not license it to anyone.

what does not belong here. no one under 18 — not as a character, not in the frame. nothing sexual. no real person who has not agreed to be there, public figures included, and no face used to suggest someone said or did something they did not. no characters, logos, or footage owned by someone else — a costumed superhero, a film still, a game character, a brand mark. nothing hateful, harassing, or promoting violence or self-harm. no identity documents, payment cards, or other people's private information. nothing illegal where you are.

generated output. models are imperfect and their filters are incomplete. a render may resemble a protected character or a real person even when you did not ask for one. if that happens, do not publish it — tell us and we will remove it. what you take out of a signal is yours to use, and yours to be responsible for, including whether you hold the rights to use it commercially where you are.

you are talking to a machine. nixm is an AI system. the character you speak with is software. it is not a person, it does not know you, and nothing it says is advice.

everything here is generated. images, video, voice and text made on nixm are produced by AI models. none of it is a recording of something that happened. if you publish a render that depicts a real person, or that a viewer could reasonably mistake for real footage, you may be legally required to label it as artificially generated — in the European Union this has been required since 2 August 2026, and other places are following. nixm does not apply that label for you. judging whether you need one, and applying it, is yours.

moderation and removal. uploads are screened automatically and may also be reviewed by a person. we may refuse, remove, or delete any upload or render, and suspend or terminate any account, at our discretion and without notice. accounts that repeatedly infringe the rights of others are terminated.

data. dialogue is ephemeral by design. no transcript accumulates to be produced; what exists is a seven-day distilled log that then dissolves. signals expire and their contents dissolve. we use standard analytics to monitor interface performance — not to harvest dialogue.

conduct. interact at your own risk. nixm^™ is a thinking interface, not a publication platform. you are responsible for what you seed and what you share.

age. nixm^™ is intended for users aged 16 and over. by entering a signal, you confirm you meet this minimum. accounts found to belong to users under 16 may be suspended without notice.

copyright — notice and takedown. if you believe material on nixm.ai infringes your copyright, write to info@turnlive.com and include: the work you say is infringed, the URL of the material, your contact details, a statement that you believe the use is not authorized, a statement made under penalty of perjury that the notice is accurate and that you are the owner or authorized to act for the owner, and your signature. we remove material promptly on a valid notice. counter-notices go to the same address.

designated agent details

our designated agent under 17 U.S.C. § 512 is the Copyright Agent, TURNLIVE LLC, 5219 Pleasure Island Road, Belle Isle, FL 32809, info@turnlive.com.

credits. credits are prepaid, non-transferable, and have no cash value. a credit is spent when a render is requested, and a spent credit is not refundable — a render you did not like is still a render. if a render fails on our side, write to info@nixm.ai and we will restore the credit. unused credits stay on your account for as long as the account is open. if you are a consumer in the European Union or the United Kingdom, you have fourteen days to withdraw from a purchase of credits; by spending a credit within that period you ask us to begin at once and give up the right to withdraw for the credits you have spent.

no warranty. the service is provided as is. we do not guarantee availability, accuracy, or that a render will match what you pictured.

liability. to the fullest extent permitted by law, the total liability of TURNLIVE LLC for any claim relating to the service is limited to the amount you paid us in the three months before the claim arose. we are not liable for indirect or consequential loss.

indemnity. you agree to cover the costs of any third-party claim brought against us arising from what you uploaded, generated, or published from a signal.

jurisdiction. by using the service from the United Kingdom, European Economic Area, Australia, or any other jurisdiction, you agree to comply with local law in addition to these terms.

conditions d'utilisation.

nixm^™ est un environnement expérimental. en amorçant un signal, vous reconnaissez le caractère transitoire des données et acceptez ces conditions.

propriété intellectuelle. la logique d'interface nixm^™, le cadre d'ancrage contextuel et l'architecture de session sont propriétaires. toute reproduction non autorisée de cette architecture est interdite.

ce que vous téléversez. vous pouvez joindre une image à un signal — comme référence pour un personnage, ou comme image à mettre en mouvement. en la téléversant, vous confirmez trois choses : vous l'avez créée ou vous en détenez les droits ; toute personne identifiable qui y figure est majeure et a donné son accord ; et elle ne contient rien de ce qui est listé ci-dessous. vous restez propriétaire de ce que vous téléversez. vous nous accordez uniquement ce que le fonctionnement du service exige — conserver le fichier, le transmettre aux fournisseurs de modèles qui produisent votre rendu, et l'afficher dans votre signal — aussi longtemps que ce signal vit, et pas au-delà. nous ne l'utilisons pas pour entraîner de modèles et nous ne le cédons à personne.

ce qui n'a pas sa place ici. aucune personne de moins de 18 ans — ni comme personnage, ni dans le cadre. rien de sexuel. aucune personne réelle qui n'a pas donné son accord, personnalités publiques comprises, et aucun visage utilisé pour laisser croire que quelqu'un a dit ou fait ce qu'il n'a pas fait. aucun personnage, logo ou extrait appartenant à autrui — super-héros costumé, photogramme de film, personnage de jeu, marque. rien de haineux, de harcelant, ni qui encourage la violence ou l'automutilation. aucun document d'identité, carte de paiement ou donnée privée d'autrui. rien d'illégal là où vous vous trouvez.

résultats générés. les modèles sont imparfaits et leurs filtres incomplets. un rendu peut ressembler à un personnage protégé ou à une personne réelle sans que vous l'ayez demandé. si cela arrive, ne le publiez pas — signalez-le-nous et nous le retirerons. ce que vous emportez d'un signal vous appartient, et vous en êtes responsable, y compris quant aux droits d'usage commercial là où vous êtes.

vous parlez à une machine. nixm est un système d'IA. le personnage avec lequel vous échangez est un logiciel. ce n'est pas une personne, il ne vous connaît pas, et rien de ce qu'il dit ne constitue un conseil.

tout ici est généré. les images, vidéos, voix et textes produits sur nixm le sont par des modèles d'IA. rien de tout cela n'est l'enregistrement d'un fait réel. si vous publiez un rendu représentant une personne réelle, ou qu'un spectateur pourrait raisonnablement prendre pour de vraies images, vous pouvez être légalement tenu de le signaler comme généré artificiellement — dans l'Union européenne, cela est exigé depuis le 2 août 2026, et d'autres juridictions suivent. nixm n'appose pas cette mention à votre place. évaluer si elle est nécessaire, et l'apposer, vous revient.

modération et retrait. les téléversements font l'objet d'un contrôle automatique et peuvent aussi être examinés par une personne. nous pouvons refuser, retirer ou supprimer tout téléversement ou rendu, et suspendre ou résilier tout compte, à notre discrétion et sans préavis. les comptes qui portent atteinte de façon répétée aux droits d'autrui sont résiliés.

données. le dialogue est éphémère par conception. aucune transcription ne s'accumule qui puisse être produite ; ce qui existe est un journal condensé de sept jours, qui se dissout ensuite. les signaux expirent et leur contenu se dissout. nous utilisons des outils d'analyse standard pour surveiller les performances de l'interface — jamais pour exploiter le dialogue.

conduite. vous interagissez à vos propres risques. nixm^™ est une interface de réflexion, pas une plateforme de publication. vous êtes responsable de ce que vous amorcez et de ce que vous partagez.

âge. nixm^™ s'adresse aux personnes âgées de 16 ans et plus. en amorçant un signal, vous confirmez atteindre cet âge minimum. tout compte appartenant à une personne de moins de 16 ans pourra être suspendu sans préavis.

droit d'auteur — notification et retrait. si vous estimez qu'un contenu présent sur nixm.ai porte atteinte à votre droit d'auteur, écrivez à info@turnlive.com en indiquant : l'œuvre concernée, l'URL du contenu, vos coordonnées, une déclaration selon laquelle l'usage n'est pas autorisé, une déclaration faite sous peine de parjure attestant l'exactitude de la notification et votre qualité de titulaire ou de mandataire, ainsi que votre signature. nous retirons rapidement tout contenu visé par une notification valable. les contre-notifications s'adressent au même endroit.

coordonnées de l'agent désigné

notre agent désigné au sens du 17 U.S.C. § 512 est le Copyright Agent, TURNLIVE LLC, 5219 Pleasure Island Road, Belle Isle, FL 32809, info@turnlive.com.

crédits. les crédits sont prépayés, non transférables et sans valeur monétaire. un crédit est consommé dès qu'un rendu est demandé, et un crédit consommé n'est pas remboursable — un rendu qui ne vous plaît pas reste un rendu. si un rendu échoue de notre fait, écrivez à info@nixm.ai et nous vous le restituerons. les crédits inutilisés restent sur votre compte tant que celui-ci est ouvert. si vous êtes un consommateur dans l'Union européenne ou au Royaume-Uni, vous disposez de quatorze jours pour renoncer à un achat de crédits ; en consommant un crédit pendant ce délai, vous nous demandez de commencer immédiatement et renoncez au droit de rétractation pour les crédits consommés.

aucune garantie. le service est fourni « en l'état ». nous ne garantissons ni sa disponibilité, ni son exactitude, ni qu'un rendu correspondra à ce que vous imaginiez.

responsabilité. dans toute la mesure permise par la loi, la responsabilité totale de TURNLIVE LLC pour toute réclamation liée au service est limitée aux sommes que vous nous avez versées durant les trois mois précédant la réclamation. nous ne sommes pas responsables des dommages indirects ou consécutifs.

garantie d'indemnisation. vous acceptez de prendre en charge les frais de toute réclamation de tiers dirigée contre nous et découlant de ce que vous avez téléversé, généré ou publié depuis un signal.

juridiction. en utilisant le service depuis le Royaume-Uni, l'Espace économique européen, l'Australie ou toute autre juridiction, vous acceptez de respecter le droit local en complément des présentes conditions.

learn

ai video & images

privacy & ephemeral ai

gifting

notes

lip sync

CAST

cinematic studio

extend

video gen

image gen

templates

nixm-live

the runner