Higgsfield takes AI video from 1 minute to over 10

Higgsfield AI unveiled GPT‑6 Astra, a model that extends video generation from one minute to over 10 minutes, allowing creators to produce longer, more ambitious projects. CEO Alex Mashrabov highlighted how the tool enables single creators to complete high-end campaigns independently by automating repetitive tasks like shot variations and 3D modeling. The company’s YouTube channel, with nearly half a million subscribers, now uses GPT‑6 Astra to generate instructional videos exceeding 10 minutes, reducing manual effort and improving consistency in longer productions.

GPT‑6 Astra addresses a key challenge in creative AI by maintaining visual consistency across extended videos, a problem previously difficult to solve. Alex Mashrabov cited its use in supporting rural schools in Kazakhstan, where the model identifies student learning gaps and generates tailored educational videos. The tool’s ability to integrate external functions, such as 3D modeling, further enhances its utility for complex creative workflows.

GPT‑Image‑2.5 introduces precise text rendering on images, a feature Alex Mashrabov describes as state-of-the-art and the first of its kind. The model enables small businesses to create promotional content with readable text and infographics without large production budgets. Higgsfield AI’s creative team collaborates with engineers to refine these tools, leveraging OpenAI’s Codex to prototype ideas efficiently.

Higgsfield AI’s collaboration between 200 creatives and 100 engineers accelerates product development, with GPT‑6 Astra reducing the gap between prototyping and production. Alex Mashrabov praised OpenAI’s advances in planning, tool calling, image generation, and voice models, noting their role in streamlining workflows. The company aims to simplify filmmaking by allowing creators to describe ideas while AI handles the intermediate steps, with voice models described as best-in-class for communication.