Compare available video generation models by capability, speed, and DA cost to pick the right one for your use case.
OpenDunes serves video generation through the same base URL and authentication you already use for text, via POST /v1/videos. The model catalog lists 16 live video models; this guide explains how to choose between them.
Different video models optimize for different trade-offs. Before picking one, answer these questions:
All live video models take a text prompt as input (text-to-video). Image-input conditioning and reference-guided generation are tracked to be built — see Turn an Image into a Video and Guide a Video with Reference Images.
Each video model's billing_unit is video_second, and its pricing.video_second_da field is the per-second rate in DA — so you can estimate the cost of a generation before sending the request.
Social media content — prioritize speed and low cost. google/veo-3.1-fast and minimax/hailuo-2.3 sit at the low end of the price range with quick turnaround.
Product demos and presentations — prioritize visual fidelity. google/veo-3.1 and openai/sora-2-pro deliver higher-end output at a higher per-second rate.
Creative / artistic — prioritize prompt expressiveness. Models like x-ai/grok-imagine-video handle stylized prompts differently; check each model's detail page and compare outputs on your own prompts.
Per-second rates run from about 24 DA/second (google/veo-3.1-fast, minimax/hailuo-2.3) through 60 DA/s (google/veo-3.1) and 91 DA/s (openai/sora-2-pro). An 8-second default clip on google/veo-3.1-fast comes to roughly 194 DA. Exact rates are in the model catalog and in pricing.video_second_da on each catalog entry.