invideo boosts color grading success rate 3x and produces 50 custom effects per day with GPT-6 Astra

Agentic video editor invideo has integrated GPT-6 Astra to significantly improve the precision and efficiency of complex video editing tasks. CEO Sanket Shah stated that Astra excels at planning edits with frame-level accuracy, completing complex tasks with fewer reasoning steps and output tokens compared to previous models.

Improved precision in color grading and editing plans

GPT-6 Astra demonstrates strength in selecting the optimal approach from various methods, including color correction, grading, regeneration, and LUT application. In particular, color grading and correction success rates improved by approximately 3x in complex scenarios requiring character tracking and separation across frames, such as changing background colors while preserving skin tones. Additionally, it can maintain the editor's original intent and execute multiple instructions simultaneously even during long tasks.

Automated generation and control of custom effects

Astra codes custom effects tailored to the video based on text descriptions and visual references, placing them on the timeline. The generated effects include controls that allow editors to adjust details. As a result, invideo editors were able to produce approximately 50 custom effects in a single day, reducing the burden of repetitive frame-level tasks while keeping creative decisions in human hands.