Higgsfield Extends AI Video Length to Over 10 Minutes Using GPT-6 Astra

Higgsfield AI has extended the maximum length of its AI-generated videos from approximately one minute to over 10 minutes by leveraging GPT-6 Astra. CEO and co-founder Alex Mashrabov explains that this capability allows smaller creative teams to execute high-end projects, such as fashion campaigns, independently. The model helps maintain visual consistency across longer durations, addressing a significant challenge in creative AI.

Enhancing Production Workflows

The platform's creative agents use GPT-6 Astra to handle repetitive tasks like generating shot variations and calling external tools for 3D modeling. This automation has streamlined content creation for Higgsfield’s own YouTube channel, which has nearly half a million subscribers. The team now produces instructional videos over 10 minutes long with reduced manual effort. Mashrabov also highlights the model's utility in educational contexts, such as creating tailored learning videos for rural schools in Kazakhstan.

Improving Image Precision and Development Speed

For small businesses, GPT-Image-2.5 provides state-of-the-art capabilities for rendering readable text and precise infographics on images, eliminating the need for large production budgets. Internally, Higgsfield uses Codex to bridge the gap between creative prototyping and production. Mashrabov notes that GPT-6 Astra can now ship production-ready code, allowing the team of 200 creatives and 100 engineers to iterate faster with less engineering friction.

Future Vision: Voice-to-Video

Looking ahead, Higgsfield aims to implement GPT-Live-1 to enable a "voice to movie" experience. This feature would allow creators to describe ideas verbally, with AI rendering the vision in real time. Mashrabov cites OpenAI’s advances in planning, tool calling, and voice understanding as key enablers for this future workflow.