Hi everyone, I'm Martin, the maker of Timeline Studio. This open-source browser video editor runs fully local-first powered by WebGPU and ONNX Runtime Web.
Most AI editing tools only simulate UI clicks to operate timelines, which is slow and unstable when facing page loading exceptions. I built a set of standardized Agent Skills to let AI directly parse and modify native timeline project files, cutting redundant browser interaction overhead.
The project supports multi-track editing, audio separation, multilingual subtitles and offline WebM export. The full MIT licensed source code will be public when we launch on August 25. I m currently optimizing the loading & inference speed of large diffusion models inside the browser, and I welcome all technical discussions and suggestions here before launch.
GitHub repo preview: https://github.com/MartinDelophy...
This is awesome! I make videos for Tiktok and Reels on Capcut which is super expensive and sometimes buggy, especially on laptop. I've begun creating a video on this product and so far it's going great! The amount of AI features is insane for what's been free for me so far. I'd say the main strong point though is how responsive everything feels, everything I click happens very quickly and so the usage flow is great. How exactly are the AI tasks done? Is there an API call to some other model?
@neuroartist Thanks so much for your thoughtful feedback! Really glad you’re enjoying Timeline Studio.
Many of our AI features run locally in‑browser via WebGPU & ONNX, that’s why you get such snappy response without waiting for remote API round‑trips. A small subset of heavy‑duty tasks can optionally use cloud resources.
Can’t wait to watch your TikTok/Reels video once it’s finished! Feel free to drop the link here, I’d love to share it. Any other suggestions are always welcome.