Disclosure: This article was written with AI assistance and reviewed by EmpirioLabs AI.
MiniMax H3 is available on EmpirioLabs. It is a multimodal video model that takes text, images, video, and audio in a single request and returns a 4 to 15 second clip at 24 fps in 768p or 2K, each with a native stereo audio track.
Four ways to drive it
H3 is one model with a mode parameter that is auto-detected from whatever you attach, so most requests do not need to set it at all.
- Text to video: a prompt on its own.
- Image to video: one image becomes the opening frame.
- First and last frame: supply an opening and a closing image and the model transitions between them.
- Reference to video: up to 9 images, 3 videos, and 3 audio clips, 12 reference files in total, to carry a subject or a style across the shot.
Prompts can run to 7000 characters and can describe action, camera work, dialogue, sound effects, and on-screen text. Aspect ratio can be set explicitly or left adaptive so it follows your input media.
How to call MiniMax H3
Video generation is asynchronous. Submit the job, then poll it:
curl https://api.empiriolabs.ai/v1/videos/generations \
-H 'Authorization: Bearer $EMPIRIOLABS_API_KEY' \
-H 'Content-Type: application/json' \
-d '{
"model": "minimax-h3",
"prompt": "A lantern drifts over a quiet harbour at dusk, slow push in, gulls calling in the distance.",
"duration": 6,
"resolution": "768p",
"aspect_ratio": "16:9"
}'The response returns a job_id and a poll_url. Poll GET /v1/jobs/{job_id} until the status is completed, and the finished clip arrives as a delivery URL in data[0].url.
To animate an existing image, add image. To transition between two frames, add last_frame as well. To carry a subject across a scene, send several images under images.
Pricing
H3 is billed per second of video at the resolution you request, with 2K priced above 768p. Two details are worth knowing before you build:
- A reference video is billed for its own duration in addition to the clip you generate, so a 5 second reference used to make a 6 second clip bills 11 seconds.
- The first 5 reference images are included, and each additional reference image is charged separately.
See the live model page and pricing page for current rates.
Start building
Try MiniMax H3 in the Playground, read the API documentation, or compare it with other models in the model catalog.
