Explore MiniMax H3 on Longtake: 4–15 second video, 768p and 2K output, first or last frames, and mixed image, video and audio references.
Listed in Longtake's current catalog. Select the model and check the available settings in the video workspace.
Longtake's H3 configuration lists 4–15 second clips at 768p or 2K. Build a focused scene—for example, a ceramic bowl rotating under changing studio light—and inspect edges and surface details in the output.
Text, first-frame, last-frame and first-and-last-frame modes are exposed. A last-frame brief can define the composition you want the scene to arrive at; describe the preceding action in the prompt.
Reference mode accepts up to nine images, three videos and three audio files. Separate appearance references from movement references so the creative brief says which parts of each source should influence the shot.
Audio files can be supplied as references. Describe the intended voice, rhythm or atmosphere, then audition the result. The current configuration does not establish a separate stereo-output control or a fixed frame-rate guarantee.
Source: https://www.longtake.ai/api/v1/model-offerings?limit=100. Checked October 16, 2026. Illustrations and scene briefs are original creative examples, not generated samples.
The current catalog exposes generation modes, but no dedicated editing mode. Do not assume every capability of an external H3 service is exposed in Longtake.
The current Longtake configuration accepts up to nine images, three videos and three audio files in reference mode. Choose sources you have permission to use and give each one a clear role in the prompt.
