What Invideo Is Reporting
Invideo positions itself as an agentic video editor, one designed to handle the constant stream of decisions that go into editing, shaping story, placing sound effects, building transitions, adjusting color, while keeping a human editor in ultimate control. According to the company, GPT-6 Astra meaningfully improves the AI agent's ability to plan and carry out complex edits.
"How Astra can plan a particular edit on a frame-level accuracy is quite stunning," the company said. Sanket Shah, invideo's CEO, added that Astra also works more efficiently than prior models invideo tested: "Astra is using far fewer reasoning sets and thus output tokens" to reach the same result, translating into faster, cheaper edit planning.
Why Color Grading Is the Hard Part
Color work is highlighted as the standout improvement, and for good reason: it requires an AI agent to choose between several overlapping techniques, correction, grading, regeneration, LUTs (look-up tables), and isolation, and apply the right one to the right part of the frame. A common example invideo cites is changing a background while preserving a person's skin tone, which requires the agent to isolate and track that person across every frame before applying color changes elsewhere in the scene.
"Earlier, we saw very high failure rates for color-grading and color-correction tasks. With Astra, the success rate improved about three times," Shah said. That threefold jump directly addresses one of the more tedious, error-prone parts of automated video editing, one where earlier models reportedly struggled significantly.
From Descriptions to Custom Effects
Beyond color work, GPT-6 Astra also helps invideo turn plain-language descriptions and visual references into working custom effects. The agent can write code for an effect tailored to specific footage, place it correctly on the timeline, and add adjustable controls so a human editor can refine it further rather than accepting a fixed result. Invideo says a small team of editors used this capability to produce around 50 custom effects in a single day, a volume that would typically require far more manual development time.
Part of a Broader Pattern
This case study follows a similar shape to other GPT-6 Astra creative and professional tool announcements OpenAI has published in recent weeks, including Harvey's legal drafting improvements and Legora's financial document review. Separately, some public demonstrations from other tools, including one from Higgsfield AI showing GPT-6 Astra grading flat Apple Log 2 footage inside DaVinci Resolve, and reporting from MindStudio describing Astra importing files, color-grading footage, and syncing clips inside Final Cut Pro, suggest color-grading competence is becoming a recurring strength OpenAI is highlighting for this model across multiple video-editing integrations, not just invideo's.
As with other OpenAI customer case studies, these figures come directly from invideo and OpenAI rather than independent, third-party benchmarking. The three-times figure describes an internal success-rate comparison against invideo's own prior results, not a standardized industry benchmark, so it's best read as a company-reported improvement rather than a verified external measurement.
