Higgsfield Promises AI Videos Over 10 Minutes Long

Higgsfield AI details how its agents and OpenAI models are used or showcased to extend videos beyond 10 minutes and accelerate software production. The company also highlights an image rendering with readable text for small organizations and an ambition for "voice-to-film." Most of the advancements mentioned are attributed to its CEO, Alex Mashrabov.

The team claims ready-to-use code and aims for voice-to-film

Alex Mashrabov asserts that GPT-6 Astra can deliver production-ready code and reduces the work needed to integrate ideas into the product. Higgsfield AI indicates it works with around 200 creatives and 100 engineers, and Alex presents Codex as empowering the creative team with the capabilities of an experienced software engineer. On the roadmap, he describes a "voice-to-film" experience where AI understands verbal intent and renders the vision in real-time, a goal the company associates with GPT-Live-1. Alex believes that advancements in planning, tool invocation, image and voice generation bring this experience closer, and he describes the voice model as "best in class." The stated ambition is to bring film creation closer to simply describing an idea.

Videos longer than 10 minutes and a tool-equipped workflow

Before GPT-6 Astra, videos were limited to about one minute; Alex Mashrabov claims that content longer than 10 minutes is now possible. Higgsfield's creative agents handle repetitive tasks and, with GPT-6 Astra, can call external tools for 3D modeling work, which, according to Alex, helps maintain consistency over longer formats. The company's YouTube channel, which reportedly has nearly 500,000 subscribers, teaches models and workflows and now relies on a workflow built with GPT-6 Astra to generate tutorials exceeding 10 minutes, where production previously required significant manual effort. Alex presents these capabilities as a lever for small teams and asserts that a single creator can lead high-end fashion campaigns from start to finish.

GPT-Image-2.5 is presented as reliable for text and infographics

For small businesses producing promotional content, Alex emphasizes the importance of readable text and accurate infographics. He claims that GPT-Image-2.5 makes this type of content accessible without large budgets. He describes it as cutting-edge and the first of its kind to reliably render text and infographics on images.

From prototype to product: organization and internal tools

Higgsfield AI states it builds its products through collaboration among about 200 creatives and 100 engineers, with creatives using Codex to transform their ideas into software prototypes. Transitioning these prototypes into production previously required significant involvement from front-end and back-end teams; Alex argues that GPT-6 Astra now reduces this workload. According to him, this decrease in friction allows the company to develop and ship better products more quickly. Alex adds that Codex gives the creative team a capability close to that of a seasoned engineer.

An educational use is being experimented with in Kazakhstan

Alex Mashrabov indicates that producing longer instructional videos also has personal significance, as he supports rural schools in Kazakhstan. He describes an initial use of GPT-6 Astra leveraging visual understanding to identify subjects where students may encounter difficulties. Based on these detections, tailored videos would be created to meet their learning needs.