Submit one request and poll it until the clip is ready. Billed by the second of video delivered, at the published rate, on the same API key as every other model.
inferway/minimax-h3-768pMiniMax H3 768P video generation with audio, served with int8 pruned weights.
Ordered in whole seconds; the delivered clip is rounded up to the model's frame grid at no extra charge.
Inferway Media Retention Policy v1: generation outputs are retained temporarily for retrieval.
t2vText to videofl2vaFirst and last frame to videoref2vaReference images to video5s124 Delivered frames6s158 Delivered frames7s175 Delivered frames8s192 Delivered frames9s226 Delivered frames10s243 Delivered frames1344×768768×1344promptseedduration_secondswidthheightfirst_framelast_frameref_image_1ref_image_2ref_image_3Ordered in whole seconds; the delivered clip is rounded up to the model's frame grid at no extra charge.
2K resolution upgrade calls the official MiniMax regeneration endpoint, billed additionally per video second at the official regeneration rate.
Context-IR prompt enhancement calls the official MiniMax endpoint, billed at official token rates for prompt input and output.
Unlike the text models, a generated clip has to be stored so you can fetch it. It is kept under the Inferway Media Retention Policy v1 for the window the catalog publishes for this SKU, then deleted. Prompts and request metadata follow the same retention as the rest of the platform.
Transparency hubGeneration is ordered through POST /v1/interactions and runs as a background job. A dropped connection never loses work already paid for: the job keeps its own id, and you poll it until the result is ready.
Transparency hub