H3 Max Hand Drawn: What the Fixed Style Actually Does

H3 Max Hand Drawn is its own fixed-style endpoint. It is not the ordinary H3 Max route with “hand drawn” appended to a prompt. The endpoint supplies the visual treatment; the request still has to explain what happens on screen and what should be heard.

Its current contract is easy to misunderstand: a text prompt is required on every request. You may also attach one image, but the image supplements the prompt rather than replacing it. The output is 768P with generated audio, and duration can be set from 5 through 15 seconds.

The generator on this site does not call the Hand Drawn endpoint. Standard text and image generations use H3 Max Turbo, while reference mode uses another H3 Max route. Our generator is useful for drafting the staging and soundtrack of a shot, but writing the preset's name there will not switch the underlying model route.

Decide whether Hand Drawn matches the job

Start with the deliverable rather than the aesthetic label.

The distinction matters for production planning. The endpoint produces a finished clip and audio, but its schema has no end-frame input, multi-image reference set, or style-strength control. If your workflow depends on changing individual artwork elements after generation, plan that work in a separate graphics or compositing tool.

Write a short scene brief, not a pile of style words

The endpoint already knows which visual preset to apply. Repeating “sketch,” “ink,” “2D,” and “hand drawn” does little to establish the event. Spend the prompt on information the model cannot infer from the endpoint name:

• who or what occupies the frame;
• where the scene takes place;
• the order of events during the available seconds;
• the shot size and camera rule;
• concrete sound cues, including any dialogue.

Think of the prompt as directions for a single take. Here is an original, untested starting point:

Inside a rooftop observatory after midnight, an exhausted archivist spreads
an antique star chart across a brass table. A drawn constellation detaches
from the paper, passes behind the lantern, then returns as the archivist
turns the page. Locked medium-wide composition from table height; the lantern
stays on the right side of frame throughout. Amber light indoors, cold blue
sky through the open dome. Audio: paper creases, quiet brass mechanisms,
three dry scratching sounds as the constellation moves, distant wind. No
dialogue and no music.

This brief supplies a timed sequence, a stable spatial rule, and sounds tied to visible sources. If a result becomes confused, simplify the sequence into earlier and later beats rather than adding more adjectives. When a recurring character drifts, reuse a compact set of concrete identifiers across attempts; the character-consistency checklist explains how to choose them.

Prompt-only and prompt-plus-image requests behave differently

In prompt-only mode, the request may specify 21:9, 16:9, 4:3, 1:1, 3:4, or 9:16. H3 Max creates the composition along with the styled scene.

In prompt-plus-image mode, the text is still required and the uploaded image is optional input. The current API documentation states that aspect_ratio is ignored when an image is present, so do not expect that field to crop the upload into a different shape. Prepare the source image at the framing you actually intend to use.

fal recommends pencil artwork for a consistent result. A source with clear foreground and background separation also gives the requested movement somewhere to occur. Treat fine marks, clothing details, and small props as material the model may reinterpret as the scene evolves—not as locked production assets.

Check the schema before promising a workflow

The current Hand Drawn API schema, checked October 14, 2026, includes:

• a required prompt;
• one optional input image;
• duration;
• aspect ratio for requests without an image;
• seed.

It does not include:

• a required final-frame image;
• several separate character references;
• an existing video for restyling;
• a negative-prompt field;
• a slider for the strength of the preset;
• source cels, paths, or transparent element passes.

The API response is a video asset with audio rather than a bundle of editable scene components. Add exact logos, captions, and title cards in post, where spelling and typography can be controlled. The on-screen text guide covers that handoff.

The current upstream price is $0.08 per second

fal currently lists H3 Max Hand Drawn at $0.08 per generated second, with audio included. The arithmetic is:

These figures are fal's public endpoint prices, not this site's credit balance or checkout price. The current schema offers no 480P Hand Drawn option. Recheck the provider page before fixing these numbers in a client quote or recurring budget.

Use a four-pass acceptance check

Judge a generation against the brief instead of the attractiveness of a paused frame:

1. Event pass: Did the actions occur in the stated order, and is the main event understandable without explanation?
2. Continuity pass: Do the character, important prop, and left-to-right positions remain coherent between the beginning and end?
3. Framing pass: Did the requested shot rule hold, or did an unrequested pan, zoom, or crop hide the event?
4. Audio pass: Do individual cues arrive with their visible causes, and does dialogue remain intelligible if dialogue was requested?

Revise one variable per attempt. Keeping the seed while changing a single sentence makes the comparison more useful than simultaneously replacing the image, duration, staging, and sound plan. Use the camera-movement diagnostic when framing is the problem, or the synchronized-audio guide when the visual sequence works but the soundtrack does not.

H3 Max Hand Drawn is a practical choice when a fixed illustrated treatment is part of the concept from the beginning. Its strongest workflow is equally specific: write the required prompt as a compact scene brief, add an image only when you need to supply the initial composition, and evaluate the resulting clip against the directions you actually gave it.