How to Use H3 Max
Choose the route that matches the shot, prepare only the inputs you need, review the credit estimate, and iterate with controlled changes.
1. Select model and generation mode
Choose Turbo for faster text or image iteration. Choose H3 Max for text, image, or multimodal reference generation. The interface updates the available controls when the model or mode changes.
2. Prepare the prompt and inputs
Describe the subject, action sequence, camera, look, sound, and important constraints. Upload a start frame, optional end frame, or authorized references only when each asset has a specific role.
3. Check duration, resolution, and credits
Select 5 to 15 seconds and either 480P or 768P. Confirm the displayed estimate before submitting; reference-to-video can add token-block charges when references exceed the included allowance.
4. Generate, review, and revise
The job runs asynchronously and appears in Video History. Review picture and sound together, identify the most important failure, and revise one variable at a time instead of rewriting the entire brief.
Questions
Frequently asked questions
Do I need to install a model?
No. The website uses hosted fal endpoints, so generation runs without a local checkpoint or ComfyUI setup.
Where do completed videos appear?
Signed-in users can review job status and completed outputs in Video History.
What happens if a generation fails?
Reserved credits are settled according to the recorded job outcome. Check the credit ledger and contact support with the generation ID if the balance appears incorrect.