Create a film trailer with AI

Your pitch feeds two Text nodes that write the shots, Nano Banana Pro makes the images, Wan 3.0 animates them, and the Montage lays in the title card and the music. Workflow visible here, no account needed.

The actor's reference board, dropped into the Media node

Interactive demo

Real workflow, real results. Pan and zoom freely.

Use this workflow

Real workflow: the pitch of an invented thriller, two shots written by Claude Sonnet 5, animated by Wan 3.0, a title card drawn by GPT Image 2.5 Sunburst and an ElevenLabs bed. The cut comes from the demo.

At a glance

You provide: a text or a brief
What the workflow produces: 3 images in 16:9 · 2 videos of 10s in 720p, 16:9 · 1 audio track, 30 s in total
Canvas structure: 13 nodes wired by 14 connections
Models wired in: Claude Sonnet 5, Nano Banana Pro, Wan 3.0, GPT Image 2.5 Sunburst, ElevenLabs Music
Cost of one full run: about 257 credits, roughly $3.08
Account: The demo is visible without an account; generating requires a free account, trial credits included

A trailer is not a small film: it is two or three images that make people want more, a title, and a bed that rises. This workflow starts from the pitch, pulls two shots that belong to the same film out of it, and returns twelve seconds cut together.

The first hard part is COHERENCE. Two shots written separately give two different films: the pitch therefore carries a visual style line, lens, palette, grain, that both prompts copy word for word, and the character shot reuses the actor sheet.

The second is the title card. A video model writes badly and an image model writes well: the title is therefore an image laid into the montage, never text asked of the video. A full run costs two hundred and fifty-seven credits, and the montage itself is free.

Actual renders from the demo

These are the files of the workflow above, exactly as the models produced them, unretouched. The starting point is shown when there is one.

Render

The start image of shot 1: the dish under the storm, one warm light at its footGenerated with Nano Banana Pro

The start image of shot 2: the astronomer's face, a warm lamp on one side, cold screens on the otherGenerated with Nano Banana Pro

Shot 1 animated: the camera moves slowly towards the dish while the storm turns behind itGenerated with Wan 3.0

Shot 2 animated: a slight push in on the face at the moment the signal answersGenerated with Wan 3.0

The title card: SIGNAL in spaced light grey capitals on black, nothing else in the frameGenerated with GPT Image 2.5 Sunburst

The thirty second bed: a low drone, a pulse that speeds up, a string swell at the endGenerated with ElevenLabs Music

The final cut: both shots, the card for two seconds, the bed underneath, twelve seconds in all

Who it is for

Directors, producers and writers who have to show a film before it is shot, or announce a short that is already finished.

Which files to provide

A three sentence pitch and one visual style line, plus a photo of your actor if a character appears.

How it works

1
Write the pitch and the style
Three sentences of story, then one visual style line: the lens, the palette, the grain, the kind of light. That line is what will hold the shots together.
2
Let the Text nodes write the shots
Each one reads the pitch and returns the prompt of ONE shot, reusing the style line. It is more reliable than writing two prompts yourself that end up drifting apart.
3
Go through a still before video
A still is fixed for nineteen credits, an animated shot costs sixty. Validate the frame and the light on the still, then animate.
4
Lay the title in as an image
The title card is a black image with one word on it, drawn by a model that can write. The Montage holds it for two seconds at the end and slides the bed underneath.

Variations and use cases

The same canvas, tuned for different needs: each variation is two or three lines of prompt away.

The teaser of a film being written

A screenplay looking for money defends itself better with twelve seconds than with a synopsis: the reader sees the light, the scale and the tone at once, and that is often what triggers the read.

Announcing a short film

A finished short needs a trailer nobody has the budget to cut. Two shots pulled from the film, a card, a bed, and the announcement exists the same day.

The pilot of a series

The same pitch gives as many shots as episodes you envisage: it is the best way to show a series holds together visually before a single scene has been shot.

The trailer of a book

For a novel rather than a screenplay, the book trailer does the same from the PDF of a chapter, with a voice-over and four shots.

The teaser of an event

A festival, an exhibition, a release: the pitch describes the event, the two shots give the mood, and the card carries the date.

Testing a tone before writing

Seeing your story in twelve seconds quickly says whether it is a thriller or a comedy. That is a decision better made before the first act.

Common mistakes

Letting each shot invent its own style

This is the mistake that gives two films instead of one. The style line, lens, palette and grain, is written once in the pitch and copied as is into every prompt.

Asking the video for the title

A video model writes approximate letters that shift from frame to frame. The title is an image, laid into the montage: it is the only way to get a clean word.

Animating before validating the still

A still is redone for nineteen credits, an animated shot for sixty. Validating the frame on the still divides the cost of trial and error by three.

Trying to tell the story

A trailer does not summarise, it makes people want. Two strong shots and a title beat six shots that explain everything and surprise nobody.

Recommended models

Frequently asked questions

Are two shots enough for a trailer?

For a twelve second teaser, yes, and it is the format festivals and social networks take best. For a full trailer, duplicate the pair of nodes as many times as you need shots.

Can I add a voice-over?

Yes: a Voice-over node written by a Text node, wired into the audio input of the Montage, exactly like the bed. The demo stops at the music to stay readable.

Does the montage spend credits?

No. The Montage node encodes in your browser: joining the shots, laying the card and mixing the music cost nothing, however many attempts you make.

How much does a trailer cost?

Two hundred and fifty-seven credits for the demo cut: three credits per written prompt, nineteen per still, sixty per animated shot, nine for the title card and eighty-four for thirty seconds of music.

Why Wan 3.0 for the shots?

Because it takes the still as a START IMAGE and keeps the frame and the character for five seconds. The shot therefore begins exactly on the image you validated.

Can I use a real actor?

Yes, with their consent: their photo goes into the Media node and serves as the reference for the shot they appear in. The demo uses the face of the founder of the site, with his consent.

How long is the cut?

Twelve seconds: two five second shots and a two second card. The Montage lets you lengthen, trim and reorder without another generation.

Is the music cleared?

It is generated for you by ElevenLabs Music and is yours under the terms of the service. That is exactly what avoids the stock bed everybody recognises.

Create a film trailer with AI

Use this workflow

Free account, no card required. The workflow opens pre-filled in your canvas.