MiniMax H3
Up to 2K video with native stereo audio, from one text, image, video and audio context — with open weights for local 768p runs
MiniMax H3 is a general-purpose multimodal video model: it reads text, images, video and audio as one context and returns 4–15 second clips with native 32 kHz stereo sound, dialogue and effects included. Its H3-Base weights are open as two checkpoints, FL2VA and Ref2VA, which render 768p on your own hardware; full 2K output runs through MiniMax's API. H3 is not yet available to generate in Imgo.
Capability snapshot
2K
maximum output, via API
768p
open weights, run locally
15s
maximum clip
See MiniMax H3 in action
A 27-second walkthrough of the workflow, the key benefits, and the results.
Core benefits
Practical value mapped to real production workflows.
Benefit
01
One context, not four pipelines
Text, image, video and audio go in as a single context, so a prompt can say "take the camera move from this clip, have the person in this image sing, match this voice" and be understood as one instruction. The reference mode takes up to 9 images, 3 video clips and 3 audio clips, 12 files in all.
Benefit
02
Sound generated with the picture
Audio is generated jointly with the video as 32 kHz stereo, so dialogue, effects and ambience arrive already aligned to the shot instead of being layered on afterwards. MiniMax lists stable dialogue in 11 languages, including English, Chinese, Japanese, Korean and Spanish.
Benefit
03
Open weights for 768p, the API for 2K
The H3-Base weights are public as two checkpoints: FL2VA for text and first/last-frame video, Ref2VA for reference-driven video. Run locally, they render 768p. The 2K pass, H3-Regenerate-2K, is not open-sourced, so full 2K output goes through MiniMax's API.
Recommended workflow
A simple three-step path from source material to a reusable result.
Step 1
Pick the mode
Use FL2VA for text-only, first-frame, last-frame or first-and-last-frame shots, and Ref2VA when the shot is built from reference images, video or audio. One run takes frames or references, never both.
Step 2
Assemble references within the limits
Up to 9 images, 3 video clips and 3 audio clips, 12 files at most. Each video or audio clip runs 2–15 s, and each type totals 15 s or less. Then say in plain words what each reference is for.
Step 3
Choose where it runs
The open H3-Base weights render 768p on your own hardware under the community license; MiniMax's API runs the full workflow up to 2K. In Imgo, H3 is not available to generate yet — draft and copy prompts from the H3 prompt guide instead.
Frequently asked questions
What resolution and length does it output?
Clips of 4 to 15 seconds at 24 fps, in aspect ratios from 21:9 to 9:16, with 32 kHz stereo audio. The open H3-Base weights render 768p; 2K comes from H3-Regenerate-2K, which MiniMax currently offers only through its API.
Does it generate audio, or do I add it later?
H3 generates audio jointly with the video: native stereo dialogue, sound effects and ambience in the same pass. In Ref2VA mode, reference audio clips can also steer voice and timing.
Is MiniMax H3 open source?
Partly. In August 2026 MiniMax released the H3-Base weights on Hugging Face as two checkpoints, FL2VA and Ref2VA, under the MiniMax H3 Community License: open weights, not an OSI open-source license. The H3-Context-IR prompt preprocessor and the H3-Regenerate-2K stage are not open, so the official 2K workflow still needs MiniMax's API.
Can I use MiniMax H3 commercially?
For self-hosted weights, the MiniMax H3 Community License applies. It does not cover the EU, the UK, the Republic of Korea or the USA (organizations there can apply to MiniMax for a license); products earning more than US$20 million a year need prior written authorization via [email protected]; commercial products must prominently display "MiniMax H3" in their user interface; and Hong Kong law governs the agreement. Using H3 through MiniMax's API is covered by MiniMax's API terms instead. This is a summary, not legal advice — read the license itself before you ship.
How many reference files can Ref2VA take?
Up to 9 images, up to 3 video clips and up to 3 audio clips, with no more than 12 files in total. Each video or audio clip must run 2 to 15 seconds, and the clips of each type must add up to 15 seconds or less.
Can I generate MiniMax H3 videos in Imgo?
Not yet. H3 is not in Imgo's live model catalog, which is why this page shows its availability as pending. You can draft and copy H3 prompts from the MiniMax H3 prompt guide today, or run the same brief on a video model that is live in Imgo.
Next step
Want to generate today?
Browse the models that are live on Imgo right now, compare their settings and credit quotes, and start from one of them.