To turn a storyboard into video with H3 Max, prepare a separate opening image for each shot, generate each clip, then cut the downloaded results together in a video editor. Uploading a sheet of six or nine panels is not a shortcut to that sequence: our image input is a first frame, not a storyboard document reader.
AI Live Generator runs H3 Max Turbo. It does not automatically separate panels, batch-render a board or assemble a film. What it can do is animate a prepared shot for 5–15 seconds, with sound directed in the prompt. Keeping those jobs separate makes it easier to replace one weak shot without losing the rest of the edit.
A board is a plan; the uploaded image is the shot
Think about what is actually inside the file you are about to upload. If it contains three pictures, gutters, arrows and a caption, those marks are part of the image. They are not reliable commands to change camera angles or move an object. The model may preserve or animate the sheet instead of producing the single scene you intended.
Use an external image editor to prepare one full-frame picture per shot. Remove panel borders, shot numbers, speech bubbles and direction arrows. Put directions in the prompt instead. Use images you own or have permission to use.
A rough drawing also makes a visual choice. If you want the film to look drawn, that may be the right input. If you want a photographed scene, prepare an approved photographic-looking opening image first; do not assume a stick-figure sketch will turn into a finished live-action shot just because the prompt says “cinematic.” This generator does not prepare those still images for you.
Use a consistent intended aspect ratio across the shots. A tiny crop from a large contact sheet may not contain enough detail for a face or prop. Enlarging it does not recover missing detail; recreate that individual frame when necessary. Our image-to-video guide explains how the first image establishes the composition.
Plan twelve seconds without requesting three-second videos
Here is an original example: an empty railway platform receives its first train of the morning. The board contains an establishing view, a lamp switching on, and the arrival. No recurring actor or spoken performance is needed to make the sequence readable.
These are proposed shot briefs, not examples we have generated or measured.
| Shot | Time in the planned edit | Requested generation |
|---|---|---|
| A: empty platform in dawn mist | 3 seconds | 5 seconds |
| B: platform lamp lights up | 2 seconds | 5 seconds |
| C: a railcar arrives and stops | 7 seconds | 8 seconds |
The edit is twelve seconds; the requests total eighteen. Our duration control starts at five seconds. You can trim a five-second file to two seconds elsewhere, but typing “two seconds” into its prompt does not lower the selected duration or charge.
Before generating, put the three stills into an external editing timeline for three, two and seven seconds. This rough timing preview costs no video-generation credits here. Does the lamp insert help the arrival, or does it interrupt it? It is easier to remove an unnecessary shot now than after paying to animate it.
The extra requested time is editing room, not guaranteed clean footage. If the useful action lands too late or a take contains errors, the planned cut may need to change.
Give each panel its own motion brief
Prepare all three opening images in the same visual treatment: a modest station, wet surfaces, overcast dawn and a restrained palette. The lamp in shot B should be visible and unlit. Shot C should show an empty platform with space for the railcar to enter. Do not upload the same wide station image for the lamp close-up and expect it to behave like an identity-only reference.
For each job, select Image to Video, upload its own first image, leave the optional last image empty and start with Balanced expansion. Set the duration in the interface to match the table. Adapt the wording below to the images you actually prepared.
Shot A — give the place a quiet beginning:
Continue from this empty railway-platform image for one five-second shot.
A thin layer of dawn mist drifts slowly along the wet rails.
Keep the camera locked and the benches and canopy still.
The platform stays empty; no train enters this shot.
Audio: soft wind through the canopy and a distant bird call.
No speech, music, writing or cut.Shot B — make one small change unmistakable:
Continue from this close-up of the unlit platform lamp.
After a brief still moment, the bulb turns on with a warm steady glow
and remains lit. Keep the lamp housing, framing and background fixed.
One five-second shot, no flicker and no camera movement.
Audio: one quiet electrical relay click as the light comes on,
with soft outdoor wind underneath. No voice, music or cut.Shot C — let the arrival finish:
Continue from this empty-platform image for one eight-second shot.
A short cream-colored railcar enters from screen left,
travels right along the track, decelerates and stops beside the platform.
Leave a short settled moment after it stops. Doors remain closed.
Fixed wide camera; preserve the platform, canopy and dawn lighting.
Audio: approaching wheels, a gentle braking hiss, then quiet engine idle.
No passengers, announcements, music, titles or cut.The prompts ask for action, not the finished montage. If a returned railcar never stops, the last shot has not fulfilled its job—even if the atmosphere looks excellent. You can simplify the action or allow more time, but neither is a guarantee.
One important input trap: the optional second image is the ending of the same generated clip, not the next storyboard shot. A wide platform view followed by a close-up lamp is an editorial cut. Generate those separately. Use a first-and-last-frame pair when you actually want a continuous transition between compatible compositions.
Budget for generated footage, not the final runtime
At this site's current rates, the three requests above cost:
- 480P: (5 + 5 + 8) × 3 = 54 credits.
- 768P: (5 + 5 + 8) × 5 = 90 credits.
One additional five-second take adds 15 or 25 credits, bringing the total to 69 or 115. A successfully generated take still costs credits if you decide it does not belong in the edit. Twelve seconds of finished film does not imply twelve seconds of billed generation.
Choose one resolution for this first sequence. If you later regenerate all three shots at another resolution, those are new paid requests, not an upgrade applied to your existing files. The cost guide covers the debit rules; Pricing explains the plans and packs that supply those credits. These numbers are our site's credits, not fal API dollars.
Review the cut before replacing a shot
Download each result separately and keep the alternatives. Simple names such as station-A-take01 and station-B-take02 are enough to avoid editing the wrong version. Store the source image, prompt and settings beside the chosen take in your own project folder.
In the editor, find the useful action inside each file rather than mechanically keeping its first three, two or seven seconds. For the lamp, keep enough unlit time to read the change and enough lit time to register it. For the arrival, do not trim away the stop just to hit the original twelve-second target.
Watch the assembled sequence without sound. Can you understand the place and event? Do the wide shots look like the same station? A timing problem may need a different cut; a canopy that changes shape may need a revised source image or another take. Replacing the lamp will not fix the platform geometry in shot C.
Then listen across both cuts. Independent clips can have different wind levels or distracting music. Balance or replace ambience in the external editor with audio you have rights to use. Keep the relay click with the lamp and the braking sound with the train; do not hide a bad cut under a sudden loud sound. Our sound-prompt guide helps with the initial brief, but soundtrack editing is not a feature of this generator.
The goal is not to animate every panel because it exists. It is to keep the shots that make the sequence work. Start with the least certain essential shot—here, the arrival—and open the H3 Max Generator when its opening image and action are ready.
Product controls and credit arithmetic checked October 2, 2026 against our generator code. The official Turbo image-to-video schema documents first/last images; provider options and this site's exposed controls are not interchangeable.