Video
MiniMax H3 Video Generator
Text or frames to video with sound — MiniMax H3 makes a short MP4, picture and soundtrack together, from a prompt and an optional first and last frame, on devices you connect.
MiniMax H3 is licensed under the MiniMax H3 Community License Agreement, Copyright (c) 2026 MiniMax. All Rights Reserved.
Model licenses
MiniMax H3 (Comfy-Org repack of FL2VA checkpoint and VAEs; GGUF-quantized by Mesh) · MiniMax-H3-Community-License
Grant is limited to the Applicable Territory (worldwide excluding the EU, UK, Republic of Korea and USA) and to the Acceptable Use Policy (Exhibit A). Written authorization from MiniMax is on file (conditional on continued compliance with commitments made to MiniMax). Display "MiniMax H3" on commercial UI; separate MiniMax authorization if yearly revenue exceeds USD 20M; no use of outputs to train other models.
Qwen3-VL-32B text encoder (as packaged in MiniMax-H3) · Apache-2.0
LightX2V MiniMax-H3 Turbo 4-step FL2VA LoRA (merged) · MiniMax-H3-Community-License
Published under the MiniMax H3 Community License (license_link LICENSE); same territorial and AUP limits as the base model.
API reference
Submit via REST
Use your personal API key (msk_…) to create jobs for this workload from your own tools. Jobs started this way run on the same devices you connected on Your Devices — at least one has to be online.
https://cp.meshlabs.cloud/v1/organizations/jobsAuthenticate with X-Organization-Api-Key: msk_your_personal_api_key. The header name matches org keys; personal keys use the same endpoint.
POST https://cp.meshlabs.cloud/v1/organizations/job-inputs/presign with the same API key, upload your file to the returned upload_url, then pass download_url in job_spec (for example as input_image_url).Request body
{
"workload_type": "minimax-h3-fl2va-gguf",
"job_spec": {
"prompt": "A photorealistic lighthouse at dusk",
"resolution": "480p",
"aspect_ratio": "16:9",
"num_frames": 124,
"keep_warm": false
}
}job_spec fields
| Field | Required | Default | Description |
|---|---|---|---|
prompt | Yes | — | What happens in the clip: the subject, the action, the setting and the camera, and what should be heard. |
input_image_url | No | — | Optional still the clip starts from. Resized to the clip's size and encoded on the node; leave empty to generate from the prompt alone. |
end_image_url | No | — | Optional still the clip ends on, with or without a first frame. Resized to the clip's size and encoded on the node. |
resolution | No | 480p | Frame size. 480p is the default; 768p is the largest. |
aspect_ratio | No | 16:9 | Frame shape. 768p offers 16:9, 9:16 and 1:1. |
num_frames | No | 124 | Clip length in frames at 24 fps: about 3, 5 or 10 seconds. The 10-second clip is offered at 480p only. |
seed | No | — | Random seed. The same seed and prompt give the same clip. |
keep_warm | No | false | Keep the model loaded on your devices for the next clip. |
cURL
curl -X POST https://cp.meshlabs.cloud/v1/organizations/jobs \
-H "X-Organization-Api-Key: msk_your_personal_api_key" \
-H "Content-Type: application/json" \
-d '{
"workload_type": "minimax-h3-fl2va-gguf",
"job_spec": {
"prompt": "A photorealistic lighthouse at dusk",
"resolution": "480p",
"aspect_ratio": "16:9",
"num_frames": 124,
"keep_warm": false
}
}'https://cp.meshlabs.cloud/v1/organizations/jobs/{job_id}Poll with the same API key until status is terminal (COMPLETED, FAILED, CANCELED, or MODERATION_BLOCKED). When content moderation is enabled, jobs may start in MODERATION_PENDING before entering the compute queue. Completed jobs include presigned download_url values on each output.
Create response
{
"id": "3fa85f64-5717-4562-b3fc-2c963f66afa6",
"workload_type": "minimax-h3-fl2va-gguf",
"status": "PENDING",
"job_spec_json": {
"prompt": "A photorealistic lighthouse at dusk",
"resolution": "480p",
"aspect_ratio": "16:9",
"num_frames": 124,
"keep_warm": false
},
"work_units": [],
"outputs": [],
"created_at": "2026-06-08T12:00:00Z",
"updated_at": "2026-06-08T12:00:00Z"
}Completed response
{
"id": "3fa85f64-5717-4562-b3fc-2c963f66afa6",
"workload_type": "minimax-h3-fl2va-gguf",
"status": "COMPLETED",
"job_spec_json": {
"prompt": "A photorealistic lighthouse at dusk",
"resolution": "480p",
"aspect_ratio": "16:9",
"num_frames": 124,
"keep_warm": false
},
"work_units": [
{
"id": "…",
"index": 0,
"status": "COMPLETED"
}
],
"outputs": [
{
"id": "…",
"output_type": "<output type>",
"filename": "<output filename>",
"size_bytes": 524288,
"download_url": "<presigned output download URL>"
}
],
"created_at": "2026-06-08T12:00:00Z",
"updated_at": "<completion timestamp>"
}Text to video with sound, from a sentence. MiniMax H3 is MiniMax's video model; this workload runs its fast four-step version, which writes the picture and a stereo soundtrack together. Give it a first frame, a last frame or both and the clip starts and ends on them. You get an MP4 at 24 frames per second — about 3, 5 or 10 seconds at 480p, and about 3 or 5 seconds at 768p — in landscape, portrait or square. It runs on the computers you connect to MESH, not on a rented GPU. The schedule is fixed — there is no guidance scale, step count or negative prompt to tune, and the soundtrack is always generated — and 4K and clips longer than 243 frames are outside what this workload serves. Every clip is labelled MiniMax H3, as the model's licence requires.