Higgsfield expands AI video generation beyond 10 minutes

Higgsfield AI says GPT-6 Astra has extended the length of videos its users can create from roughly one minute to more than 10 minutes. The startup, which builds AI tools for video and image creation, is using the model alongside GPT-Image-2.5, Codex and a prospective voice-to-video workflow based on GPT-Live-1.

CEO and co-founder Alex Mashrabov said the longer format allows smaller creative teams to attempt work that previously required more production capacity. He said a single creator can now execute a high-end fashion campaign from start to finish with models including GPT-6 Astra and GPT-Image-2.5.

Longer generation workflow targets consistency

Higgsfield’s creative agents already handle repetitive production work, such as generating multiple variants of a shot. With GPT-6 Astra, those agents can call external tools for complex manual tasks in 3D modelling. Mashrabov said this capability helps tackle the challenge of maintaining consistency over a longer video.

The company has applied that approach to its own YouTube channel, which Mashrabov said has almost half a million subscribers and teaches creators how to use AI models and workflows. Producing the material formerly involved considerable manual effort. Higgsfield now uses an AI workflow to generate instructional videos lasting more than 10 minutes.

The longer-video capability also supports an education experiment for rural schools in Kazakhstan. Mashrabov said Higgsfield is using GPT-6 Astra’s visual understanding to identify subjects students may find difficult and to create videos tailored to their learning needs.

Images and software prototyping

For smaller businesses, Higgsfield positions GPT-Image-2.5 as a tool for promotional content where readable text and precise infographics are essential. Mashrabov said the model can reliably render text and infographics on top of imagery, potentially reducing the need for large production budgets for those assets.

Higgsfield’s product work involves about 200 creatives and 100 engineers. The company says Codex lets creatives turn ideas into software prototypes, while GPT-6 Astra reduces the engineering work required to move those prototypes into production. That operational focus follows Higgsfield’s $400 million funding round and its $400 million funding round at a $5.4 billion valuation, as the company seeks to convert creative workflows into product capabilities.

Voice-to-video remains the next goal

Higgsfield’s next objective is what Mashrabov calls “voice to movie”: a creator describes an idea aloud while AI understands the intent and renders the result in real time. He linked that prospect to advances in planning, tool calling, image generation and voice, including GPT-Live-1.

For businesses, the immediate implication is to evaluate AI video tools around specific production bottlenecks: repeatable shot work, 3D tasks, instructional content and the handoff from prototype to production. Longer outputs may be useful only where teams can retain review processes and ensure the resulting material meets their required standard.