MiniMax H3 Highspeed

MiniMax H3 Highspeed is a faster version of minimax/h3, the multimodal video model from MiniMax. It balances quality and speed, which makes it a good fit for everyday content creation where you don’t need 2K output.

What it does

• Text-to-video: describe a scene and get a video back.
• Image-to-video: give it a first frame, a last frame, or both, to control how the video starts and ends.
• Reference generation: give it reference images, videos, or audio to guide the character, motion, camera, style, voice, or editing rhythm.

Videos can be 4 to 15 seconds long, at 768P.

How it compares to the other H3 models

See minimax/h3-lite and minimax/h3.

Inputs

• prompt: the text describing the video.
• first_frame_image: optional first frame. When you provide one, the video uses the image’s aspect ratio.
• last_frame_image: optional last frame.
• reference_image_urls: optional public URLs for up to 9 reference images.
• reference_video_urls: optional public URLs for up to 3 reference videos.
• reference_audio_urls: optional public URLs for up to 3 reference audio clips. Reference audio needs to be combined with a reference image or video.
• duration: video length in seconds, from 4 to 15.
• ratio: aspect ratio for text-to-video and reference generation. Image-to-video uses the input image’s aspect ratio.