Feedback
AI Ad Video Example
Loading...
minimax h3 video model
Type a prompt and let the MiniMax H3 video model render 2K clips with synced stereo sound — text, image, and audio inputs in one pass.
All Tools
Discover our comprehensive AI-powered animation toolkit
Inside the MiniMax H3 Video Model
Built by MiniMax and hosted on fal.ai from day one, this open-weight omni model reads text, stills, clips, and sound together inside one context. It renders 2K video carrying its own stereo soundtrack for up to 15 seconds, allows pinpoint edits to a chosen region, keeps on-screen text crisp, and accepts as many as 12 reference files per run.
- Every Input in a Single PassOne run can mix as many as 9 stills, 3 clips, and 3 audio tracks, so character identity, performance, camera work, and sound stay aligned instead of drifting apart.
- Stereo Sound Baked InOutputs arrive with original score, spoken lines, foley, and room tone already matched to the cut, and a voice can be transferred or cloned from a reference recording.
- Surgical Region EditingSwap a product, repaint a sign, redub a line, or flip a scene from day to night — only the area you mark changes while everything else in the frame holds steady.
Calling the MiniMax H3 Video Model API in Three Moves
Three short steps carry you from an empty project to a finished 2K clip with matched audio.
Core Capabilities of the MiniMax H3 Video Model
Three endpoints, one shared multimodal context, built-in stereo sound, region-level editing, crisp on-screen typography, and usage-based billing — everything needed for a full 2K pipeline through fal.ai.
Three Routes to a Clip
Text-to-video, image-to-video with first and last frame control, and reference-to-video — pick whichever fits the shot you already have in mind.
Twelve Reference Slots
Load 9 stills, 3 clips, and 3 audio tracks; identity, performance, camera movement, composition, and cutting rhythm are all read from them.
Crisp Text and Live UI
End cards, captions, and brand marks come out legible, while real interfaces — landing pages, game menus, HUDs, animated type — can be brought to life.
Room for a Full Shot List
Send an entire sequence description in one request; prompts of up to 7,000 characters keep every beat of the scene under your control.
2K Output at 24fps
Deliver 2K video with a 1440px short edge, running as long as 15 seconds, in six aspect ratios plus an adaptive option.
Billing That Follows Usage
Serverless, pay-per-use pricing with no minimums and no subscriptions, plus commercial rights over whatever you generate.
MiniMax H3 Video Model: Common Questions
Answers to the questions people ask most about running the MiniMax H3 video model through fal.ai.
What exactly is the MiniMax H3 video model?
An open-weight, general-purpose omni-modal generator from MiniMax, offered on fal.ai as a Day 0 ecosystem partner. It handles text, stills, clips, and audio within one context and produces 2K video carrying its own stereo soundtrack, up to 15 seconds long.
Which endpoints can I call?
Three of them: text-to-video, image-to-video with optional first and last frame control, and reference-to-video, which pins down subjects, styles, motion, camera work, and voices from the material you supply.
What output settings are available?
2K resolution with a 1440px short edge at 24fps, running from 5 to 15 seconds, in 21:9, 16:9, 4:3, 1:1, 3:4, and 9:16 plus an adaptive mode.
Does audio come out of the box?
Yes. Each run returns stereo sound — original score, dialogue, foley, and ambience — aligned to the cut, with the option to transfer or clone a voice from a reference recording.
How many reference files are allowed?
Twelve in total: 9 images, 3 video clips of 2-15 seconds each, and 3 audio tracks of the same length. Audio has to be paired with at least one image or clip.
Are commercial rights included?
Yes. Content produced through the fal.ai API can be used in commercial projects, under the usage terms fal.ai sets out.
Put the MiniMax H3 Video Model to Work
Describe your scene, attach references, and pull a 2K clip with matching stereo audio in one request — multimodal input, region-level edits, and usage-based pricing on fal.ai.