Guozhen AIGlobal AI field notes and model intelligence

Realtime AI News

Higgsfield Pushes AI Video From 1 Minute to Over 10 With OpenAI Models

OpenAI says Higgsfield is using GPT-6 Astra to extend creators' AI video from 1 minute to over 10, while GPT-Image-2.5 brings more precise visuals. The company also points to GPT-Live as a path toward voice-driven filmmaking.

Published
Higgsfield借助OpenAI新模型把AI视频时长从1分钟拉到10分钟以上
Image source: higgsfield.ai

Higgsfield is taking AI video from 1 minute to over 10, according to OpenAI, with GPT-6 Astra doing the heavy lifting on longer generation.

Alongside the longer timelines, the company uses GPT-Image-2.5 for more precise visuals, addressing the control problems that have kept generated imagery from being reliably usable.

OpenAI also highlights GPT-Live, which it describes as a path to voice filmmaking.

Taken together, the updates point in one direction: creators are no longer limited to a few seconds of footage or a single frame, but can organise longer material around voice-driven instructions.

Duration has been one of the hard limits on commercial AI video. Moving from 1 minute to over 10 changes the question from whether the output is usable at all to how it fits into a finished piece.

What to watch next: how widely these capabilities are opened up, how they are priced, and what product form the voice-based filmmaking path eventually takes.

Why it matters

Longer clips and tighter visual control move AI video closer to supporting complete narrative work, which would reshape both editing tools and creator workflows.

OpenAIAI Video
Back to realtime news

Nearby Updates

All