Show HN: A pure TypeScript library for editing videos in the Browser

3 points

a year ago

Hey HN,

We’re building Diffusion Studio (YC F24) – an AI-native, client-side video editing engine that replaces the need for FFmpeg and enables advanced video editing in the browser.

The Problem

Video processing is still extremely difficult. FFmpeg (written in 2000) remains the default toolkit, despite lacking real-time support, modern text rendering, and animation capabilities. Most web-based editors (e.g. https://www.veed.io/) rely on server-side rendering, which adds complexity and delays due to footage uploads and queuing rendering pipelines.

In March 2023, I was automating my own social media content and realized FFmpeg/Moviepy wasn’t enough for modern workflows. That’s when I turned to Chromium, which had just introduced a low-level encoding API capable of real-time, high-performance video rendering (https://developer.mozilla.org/en-US/docs/Web/API/WebCodecs_A...). Seeing the potential, I quit my job and spent the last two years building a video processing engine designed for the AI era—fully client-side and optimized for automation.

What’s Special?

Diffusion Studio is essentially a headless CapCut for developers. Key features:

  - Fully client-side – no WASM-compiled FFmpeg, no server required
  - Keyframes, captions, and layering – built-in advanced editing capabilities
  - Blazing fast – benchmarks show it outperforms DaVinci Resolve for video export/rendering on an M1 MacBook Pro
  - AI-ready – optimized for programmatic editing and content generation
  - Great for backend rendering – Works with Playwright or Puppeteer, making it a go-to choice over FFmpeg for many server-side workflows

Mission & Milestones

Our mission is to enable anyone—whether it’s AI models or junior developers—to leverage our high-level engine to build powerful video tools and scalable editing pipelines for large-scale content creation.

So far:

  - A YC startup from the current batch (W25) is using Diffusion Studio to build video editing agents, due to its agentic-first architecture
  - A leading AI research lab is integrating the engine into its creative playground to edit and refine generative content
  - Several edtech startups are using Diffusion Studio to transform study materials into dynamic explainer videos

If you’re building tools for video-heavy platforms or looking for a better alternative to FFmpeg, we’d love to hear your thoughts.

Try It Out here: https://www.npmjs.com/package/@diffusionstudio/core

Looking forward to your feedback!

- Fully client-side – no WASM-compiled FFmpeg, no server required - Keyframes, captions, and layering – built-in advanced editing capabilities - Blazing fast – benchmarks show it outperforms DaVinci Resolve for video export/rendering on an M1 MacBook Pro - AI-ready – optimized for programmatic editing and content generation - Great for backend rendering – Works with Playwright or Puppeteer, making it a go-to choice over FFmpeg for many server-side workflows

- A YC startup from the current batch (W25) is using Diffusion Studio to build video editing agents, due to its agentic-first architecture - A leading AI research lab is integrating the engine into its creative playground to edit and refine generative content - Several edtech startups are using Diffusion Studio to transform study materials into dynamic explainer videos