what sign up
nixm nixm

← learn

AI book trailer from your book cover — what one render really is

October 26, 2026

You can hand an AI video tool a book cover and a blurb and ask for a trailer. What comes back is a script and one shot, not a film. Below is a real attempt on nixm: what was asked, what rendered, where the cover went, and what we changed so the tool tells you that up front. Then the honest way to build a book trailer with AI, one shot at a time.

What was asked

A tester uploaded a book cover, pasted the back-cover blurb, and asked for a trailer. nixm wrote a full 60-second vertical trailer: a cold open, a scene breakdown, visual direction for each beat, a closing card. The script was good. He picked the cold open and asked for it vertical.

What rendered

One clip. Eight seconds. 1080×1920. A hand holding a phone in a dark kitchen, the lock screen lighting up. With sound. Technically exactly what the cold open described.

Two things were wrong with it as a book trailer. It was 8 seconds of a 60-second plan, and the book was nowhere in it. The cover sat in the signal as a reference and never reached the frame.

Nothing failed. The tool planned a film and delivered a shot, and nobody said so.

Why that happens

AI video models render one short piece at a time. On nixm a plain clip is one continuous shot of about 8 seconds. A cast scene with a character runs 3 to 30 seconds and can cut between several shots inside one render. None of the video models we know of renders a 60-second script in one pass: each render is generated on its own, with its own lighting, its own actor and its own sound, and a trailer is put together from several.

A stored image is a separate problem. A plain text-to-video clip carries no reference image at all. The model invents the scene from words. So "a kitchen at night" renders a kitchen, and your cover stays in the drawer. A cover gets on screen in exactly three ways:

  • Animate it. The cover becomes the first frame of an 8-second clip and the camera moves on it. This is image-to-video, the same path as turning any photo into a clip.
  • Make it a scene on its own. The cover is bound to the scene engine with no character in the shot: its layers come apart and move, the camera pushes in, and the title and author stay legible. 3 to 30 seconds, several beats cut inside one render, at 1 video credit per 5 seconds.
  • Cast it as a prop. Put a character in the shot and give them the book. The scene engine binds the cover image to the object in their hands.

None of those is "render my trailer". Each is one render.

What we changed

The failure was not the render; it was the promise. So nixm now knows its own production limits and says them before you spend anything. Ask it for a trailer and it still writes the script, as long as you want it. Then it tells you what one render is here: one shot, about 8 seconds, one video credit for a clip, and which beat it would shoot first. The rest stays on the page as script until you ask for the next shot.

It is a prompt-level change, so it raises the odds rather than guaranteeing the sentence. In our tests, nixm planned shot by shot and rendered nothing unasked on all six trailer-shaped requests, named the first shot on four, and stated a shot's length on two. Ordinary clip requests read as before.

The cover path got two fixes of its own. A portrait cover used to animate as a 16:9 clip with the title cropped out of every frame; it now keeps its own orientation, so the title and author stay in the shot. And nixm can now aim an animation at a stored reference by name, so "animate the cover, slow push in" does what it says instead of animating whatever still came last.

First frame of the cover animation before the fix: the cover centre-cropped to landscape, title cut off.
Before: the cover as first frame, cropped to 16:9. The tagline and the door survive; the title and author do not.
First frame of the cover animation after the fix: the portrait cover with title and author visible.
After: the same cover, in its own orientation. Title and author are in the first frame; a 2:3 jacket in a 9:16 frame loses a sliver at each side.
The cover, animated: 8 seconds, portrait, with sound. One video credit. Nothing was edited afterwards.

How to build a book trailer with AI, honestly

A trailer is a sequence of shots, so build it as one.

  1. Write the script first, then cut it to three or four shots. A plain shot is 8 seconds. A 30-second trailer is four of them. A 60-second one is seven or eight and will cost accordingly. Most book trailers that work on social are 15 to 30 seconds.
  2. Open or close on the cover. Animate the cover for the title shot. One camera move, a push in or a slow drift, nothing clever. Eight seconds of a cover with sound is a stronger title card than any text overlay.
  3. Shoot the mood beats as plain clips. The kitchen, the phone, the empty street. These carry no reference, so describe them tightly and let the model invent. One video credit each.
  4. If your book has a face, cast it. A protagonist, or you as the author. A cast scene holds a character across shots and can put the book in their hands as a prop. 3 to 30 seconds at 1 credit per 5, vertical for social, with several beats cut inside one scene and the character speaking in sync. How continuity actually behaves is in making a short film with AI.
  5. Extend rather than re-shoot when a shot is going somewhere. EXTEND adds about seven seconds to the same shot for one credit, sound carried through, up to four passes per clip.
  6. Join the shots. On nixm, STITCH puts 2 to 8 takes of the signal into one video, in the order you pick, with hard cuts, up to 120 seconds: 1 video credit per 30 seconds, one more for an instrumental score under it. For title cards, transitions or your own music, download the shots and assemble them in any editor. This is the step that makes it a trailer rather than a clip.

What it cost

The failed attempt: one still (2 chat credits) and one HD clip (1 video credit). The cover animation in the test: 1 video credit, 8 seconds, with sound. A full 30-second trailer built as above is four clips, about 4 video credits, plus 1 to stitch them into one video: roughly CA$20 on the 5-video pack, before retakes. The script costs one message.

Try your cover

A new nixm account includes one free HD clip with sound. Open the cinematic studio, upload your cover, and ask for the title shot first.

frequently asked

Can AI make a book trailer from a book cover?

It can write the trailer and render it one shot at a time. The cover gets on screen by animating it as the first frame of a clip, or by casting a character who holds it. No tool we know of renders a whole 60-second trailer in one pass.

How long is one AI video clip?

About 8 seconds for a plain clip on nixm. Cast scenes run 3 to 30 seconds. Extending a clip adds about seven seconds per pass.

Why isn't my book cover in the video?

A plain text-to-video clip carries no reference image; the model invents the scene from words. Animate the cover itself, or cast it as a prop in a scene with a character.

Can the trailer be vertical?

Yes. Plain clips, animations and cast scenes can all be portrait. A portrait cover animates in portrait and keeps its title in frame.

What does a book trailer cost on nixm?

One video credit per 8-second clip or animation, 1 per 5 seconds of cast scene. A 30-second trailer built from four shots is about 4 video credits, plus 1 to stitch them into one video. The script is one message.

Can I try it free?

Yes. A new nixm account includes one free HD video with sound, no card required.

cast one character. put it anywhere.

try it now · cinematic studio

cast one character. put it anywhere.

cinematic ai video, with sound. nixm proposes every scene before a credit is spent.

start directing → first video free · no subscription

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
or