Higgsfield AI users can now create videos longer than 10 minutes with GPT‑6 Astra, according to CEO and co-founder Alex Mashrabov, extending a previous ceiling of around one minute. The company has put that capability into a workflow for generating instructional videos, which the announcement says also run past 10 minutes.
Higgsfield AI’s creative agents produce alternative versions of shots and handle other repetitive production work. GPT‑6 Astra also lets those agents use outside tools for complicated 3D modeling tasks, which Mashrabov says helps preserve visual consistency as videos get longer. He says GPT‑6 Astra and GPT‑Image‑2.5 can enable an individual creator to carry a high-end fashion campaign through its entire production process.
The company’s YouTube channel provides tutorials on AI models and creative workflows and has almost half a million subscribers, according to Mashrabov. Making those tutorials previously involved substantial hands-on work, before the team developed its GPT‑6 Astra workflow. Mashrabov also supports rural schools in Kazakhstan and describes an early project that uses the model’s visual understanding to flag topics where students might be struggling and build videos aimed at those needs.
For small businesses, Mashrabov points to GPT‑Image‑2.5’s ability to place readable words and precise infographics within images for promotional material. He says that capability serves businesses with limited production budgets. Higgsfield AI develops its product with roughly 200 creatives working alongside 100 engineers.
Its creatives use Codex to build software prototypes from their ideas. Moving those prototypes into production previously depended heavily on engineers working on both the front end and back end. Mashrabov says GPT‑6 Astra cuts the effort involved in incorporating those ideas into the product and helps the company ship improvements faster.
Longer video limits have also featured in camera updates: Meta’s version 6.0 update in 2024 raised Ray-Ban Meta smart glasses recording from one minute to three minutes. That change required adjusting camera settings because the default remained 60 seconds, and longer recordings consumed battery power faster.
Higgsfield AI’s next ambition involves GPT‑Live and filmmaking directed through speech. Mashrabov envisions creators describing what they want aloud while AI interprets their intentions and renders the film in real time. He sees OpenAI’s work on voice, planning, image generation and tool use as progress toward that experience.
