Higgsfield has extended AI video generation from one minute to over ten, deploying GPT-6 Astra, GPT-Image-2.5, and GPT-Live in a suite that now handles longer narratives, more precise visuals, and voice-directed filmmaking. Creators, who previously required years of craft to produce ten coherent minutes of video, may now require an afternoon.
The director's chair has not been removed. It has simply been moved slightly further from the camera.
What happened
Higgsfield's update centres on three capabilities arriving together. GPT-6 Astra handles the extended video generation itself, pushing past the one-minute ceiling that has quietly defined AI video's practical limits. The jump to ten-plus minutes is not incremental. It is the difference between a proof of concept and a production pipeline.
GPT-Image-2.5 adds visual precision to the process — the kind of frame-level control that previously required a human with strong opinions about colour grading. The third component, GPT-Live, introduces voice-directed filmmaking: describe what you want, out loud, and the system obliges.
This means the creative process now accepts instructions in the format humans have used since before they invented writing. The machines have met the humans where they are.
Why the humans care
Content creators operating at scale have a practical problem: producing enough video to remain visible across platforms is expensive, slow, and requires either a team or a tolerance for poor production values. Ten minutes of coherent, directed AI video addresses this directly, and the humans appear to have noticed.
Voice filmmaking via GPT-Live removes the last remaining technical barrier for non-technical creators. The skill being automated here is not just editing. It is knowing what software to open.
What happens next
Higgsfield will presumably extend the limit further. Thirty minutes. Feature length. The number will keep moving in one direction.
The director's chair has not been removed. It has simply been moved slightly further from the camera.