HiDream.ai's HiDream V1 Video Model Debuts at 1080p With Synced Audio

HiDream.ai has launched HiDream-O1-Video-1.0, or HiDream V1, a native omnimodal video model that outputs 1080p clips of 5 to 20 seconds with audio generated alongside the picture rather than added afterward.

The launch came in a September 17 announcement from the Beijing company, also carried by Asia News Network. HiDream V1 takes text, images or video as input and builds dialogue, sound effects and picture together in one pass rather than adding sound to finished footage.

Content-adaptive duration stands out. Rather than forcing a scene into a length the user typed in, HiDream V1 picks a clip's length itself, within the 5-to-20-second window, based on how the action plays out — a short gesture gets a short clip, a busier scene gets more seconds. That comes from HiDream.ai's "understand, plan, generate" framework: before any pixels appear, the model plans shot length, setting, character state, movement, camera motion, dialogue and ambient sound, then generates visuals, motion and audio together, before a reward model checks physical plausibility and sync.

HiDream V1 completes a four-model lineup — Image, Video, World and Embodied — built on a shared architecture HiDream.ai calls UiT. Its sibling HiDream-O1-Embodied launched September 8 and topped RoboColiseum's robustness category with a score of 0.692, per that release.

HiDream.ai says HiDream V1 debuted at No. 4 on the Artificial Analysis Image to Video Leaderboard (With Audio) and No. 8 on the Arena.ai Image-to-Video leaderboard, both with a standing disclaimer that rankings can move. Checked directly on September 20, 2026, both boards still show it: Artificial Analysis lists "HiDream-O1-Video" at rank 4 (Elo 1176), and Arena.ai lists "hidream-o1-video-1.0" at rank 8 (score 1452±12), flagged as "Preliminary." Independent outlet ifeng.com reported the same two placements after testing the model itself.

The launch also came with a Series C+ round backed by New Micro Capital, Jiaozi Capital and ICBC Capital — following a July RMB 1.5 billion Series C — earmarked for model research, computing infrastructure and product rollout. "What matters is whether a model can genuinely understand a creator's intent and how objects, actions and sounds interact in the real world," HiDream.ai chief technology officer Yao Ting said in the release.

HiDream V1 is not available to creators yet — HiDream.ai says it is currently in internal testing only, with general availability promised "in the near future." Updates are expected at hiharness.ai.