HiDream-O1-Video API ships image-to-video only

HiDream-O1-Video-1.0 ships through HiDream.ai's public API as a single-image-to-video model, a narrower thing than the natively omnimodal system the Beijing company announced in September.

Documentation on HiDream's HiHarness platform describes HiDream-O1-Video-1.0 as "a single-image-to-video model that generates a video from one reference image and a text prompt." The model ID is fixed at HiDream-O1-Video-1.0.

The generation endpoint takes one required image field and an optional prompt. The remaining controls are force_10s, a toggle for forcing exactly ten seconds, and a client-defined request_id. There is no text-to-video mode, no video input, no multi-reference support, no resolution selection and no audio switch. Output resolution preserves the aspect ratio of the input image, so the model returns whatever shape the source image sets.

That is a smaller surface than the September 17 announcement, which introduced HiDream-O1-Video-1.0, or HiDream V1, as a "native omnimodal" model supporting text, images and video as inputs and generating 1080p clips of 5 to 20 seconds with natively synchronized audio. HiDream.ai, the maker, was founded by Dr. Mei Tao, its founder and CEO; Yao Ting is its chief technology officer. Quantslant reported that the same release described the model as in internal testing at announcement.

The rankings tell the narrower story too. The launch release placed HiDream V1 fourth on the Artificial Analysis Image to Video Leaderboard (With Audio). By October 11, reads by Quantslant and OrcaRouter put it sixth on the Artificial Analysis image-to-video leaderboard at 1,175 Elo, at a benchmarked $5.80 per minute of generated video. It does not appear on the text-to-video board, which fits the missing mode in the API.

Rollout messaging has been inconsistent. On September 30 the official @vivago_ai account posted that HiDream-O1 Video was "now available in vivago R1 Studio," as Quantslant noted. On October 9, HiDream announced the global availability of vivago R1, an agent for videos up to five minutes, and said HiDream-O1-Video-1.0 would join R1 in a later update. HiDream has not clarified which holds.

For creators, the gap changes the workflow: a starting frame is required, so images come first. The model is not listed on fal.ai, Replicate or OpenRouter, leaving the HiHarness API, and HiDream publishes no official price list.

To call HiDream-O1-Video-1.0, get a HiHarness API key and POST one reference image to /api/maas/gw/v1/videos/generations; vivago R1 is available globally at vivago.ai.