A longer workflow, not a single ten-minute generation
An OpenAI customer story published October 13, 2026 describes Higgsfield's image and video work. Higgsfield's CEO says a workflow using GPT-6 Astra enables instructional videos longer than ten minutes, beyond an earlier roughly one-minute limit. This is a customer-reported production outcome, not a specification for one video model generating ten uninterrupted minutes.
The account emphasizes agents connecting repetitive shot variations, external tools, and complex 3D work to address consistency. It does not provide a matched controlled comparison, final acceptance rate, retry count, or full production cost. It should therefore not be treated as independent benchmarking or guaranteed availability for every account.
Voice-driven filmmaking is a future ambition
The story's voice-to-movie experience is described as a next step, so current functionality and future plans need to remain distinct. Creators should examine character and prop continuity, editable segments, human editing time, and cost per finished result rather than duration alone. The broader significance is a shift from isolated clip generation toward workflows that connect planning, tools, and editing.