IT Home On October 14, OpenAI announced in a statement today (October 14) that it has reached a new milestone in its collaboration with the AI video generation platform Higgsfield. After integrating GPT-6 Astra, the maximum duration of AI videos generated by Higgsfield will be increased from 1 minute to 10 minutes.
The IT Home provides the official introduction video as follows:
GPT-6 Astra supports creative AI agents in calling external tools to handle complex tasks such as 3D modeling. According to Higgsfield, this capability is used to maintain consistency in characters, scenes, and visual details of long videos, and it is a key challenge in long video generation.
Higgsfield has used relevant processes on its YouTube channel. The channel has nearly 500,000 subscribers, and the team uses GPT-6 Astra to generate teaching videos lasting over 10 minutes, reducing the extensive manual work that was needed before.
In terms of images, Higgsfield integrates GPT-Image-2.5. The company claims that this model can render readable text and infographics stably on images, helping small businesses generate promotional materials at a lower production cost.
Higgsfield also uses OpenAI’s Codex to accelerate product development. The company has approximately 200 creative professionals and 100 engineers. The creative professionals can use Codex to transform ideas into software prototypes, reducing the need for involvement from front-end and back-end engineers.
Higgsfield also revealed the next phase of direction: exploring “speech to video” creation based on GPT-Live-1. Users can verbally describe their ideas, and the AI will understand the speech intent and generate video content in real time.