@diffusionstudio/core

Diffusion Studio

Getting Started

Diffusion Studio is an open-source, browser-based video editing library that allows developers to automate video editing workflows at scale, build custom editing applications, or seamlessly integrate video processing capabilities into existing projects.

Documentation

Visit https://docs.diffusion.studio to view the full documentation.

Why Use Diffusion Studio

💻 100% client-side
📦 Fully extensible with Pixi.js
🩸 Blazingly fast WebGPU/WebGL renderer
🏎️ Cutting edge WebCodecs export

Getting Started

npm install @diffusionstudio/core

Basic Usage

Let's take a look at an example:

import * as core from '@diffusionstudio/core';

const source = await core.VideoSource // convenience function for fetch -> blob -> file
  .from('https://diffusion-studio-public.s3.eu-central-1.amazonaws.com/videos/big_buck_bunny_1080p_30fps.mp4');

// create a video clip and trim it
const video = new core.VideoClip(source) // compatible with the File API
  .subclip(0, 160); // The base unit is frames at 30 FPS

// create a text clip and add styles
const text = new core.TextClip({ 
  text: 'Bunny - Our Brave Hero', 
  position: 'center', 
  stop: 80, 
  stroke: { color: '#000000' } 
});

const composition = new core.Composition(); // 1920x1080

// this is how to compose your clips
await composition.add(video);  // convenience function for 
await composition.add(text);   // clip -> track -> composition

// render video using webcodecs at 25 FPS
// use resolution: 2 to render at 4k 
new core.Encoder(composition, { fps: 25 }).render();

This may look familiar to some. That is because the API is heavily inspired by Moviepy and Swift UI. It models the structure of popular video editing applications such as Adobe Premiere or CapCut. The current state can be visualized as follows:

Whereas each track contains zero or more clips of a single type in ascending chronological order.

A track will be created implicitly with composition.add(clip) however you can also create them manually like this:

const track = composition.createTrack('text');
await track.add(text0);
await track.add(text1);
await track.add(text2);
...

Examples

You can find more examples here., or give them a whirl on: https://examples.diffusion.studio

https://github.com/user-attachments/assets/7a943407-e916-4d9f-b46a-3163dbff44c3

How does Diffusion Studio compare to Remotion and Motion Canvas?

Remotion is a React-based video creation tool that transforms the entire DOM into videos. It's particularly suited for beginners, as web developers can start creating videos using the skills they already have.

Motion Canvas is intended as a standalone editor for creating production-quality animations. It features a unique imperative API that adds elements to the timeline procedurally, rather than relying on keyframes like traditional video editing tools. This makes Motion Canvas ideal for crafting detailed, animated videos.

In contrast, Diffusion Studio is not a framework with a visual editing interface but a video editing library that can be integrated into existing projects. It operates entirely on the client-side, eliminating the need for additional backend infrastructure. Diffusion Studio is also dedicated to supporting the latest rendering technologies, including WebGPU, WebGL, and WebCodecs. If a feature you need isn't available, you can easily extend it using Pixi.js.

Current features

• Video/Audio trim and offset
• Tracks & Layering
• Splitting clips
• Html & Image rendering
• Text with multiple styles
• Web & Local Fonts
• Custom Clips based on Pixi.js
• Filters
• Keyframe animations
  • Numbers, Degrees and Colors
  • Easing (easeIn, easeOut etc.)
  • Extrapolation 'clamp' | 'extend'
• Realtime playback
• Hardware accelerated encoding via WebCodecs
• Dynamic render resolution and framerate

Contributing

Contributions to Diffusion Studio are welcome and highly appreciated. Simply fork this respository and run:

npm install

Before checking in a pull request please verify that all unit tests are still green by running:

npm run test

Background

This project began in March 2023 with the mission of creating the "video processing toolkit for the era of AI." As someone passionate about video editing for over a decade, I saw Chrome’s release of Webcodecs and WebGPU without a feature flag as the perfect moment to build something new.

Currently, most browser-based video editors rely on server-side rendering, requiring time-consuming uploads and downloads of large video files. With Webcodecs, video processing can now be handled directly in the browser, making it faster and more efficient.

I’m excited to be part of the next generation of video editing technology.

Compatability

✅ Supported 🧪 Experimental ❌ Not supported

Desktop

Browser		Operating System
Chrome	✅	Windows	✅
Edge	✅	Macos	✅
Firefox	✅	Linux	✅
Safari	✅
Opera	✅
Brave	✅
Vivaldi	✅

Mobile

Browser		Operating System
Brave Android	✅	Android	✅
Chrome Android	✅	iOS	🧪
Firefox Android	🧪
Opera Android	✅
Safari iOS	🧪

	Demultiplexing	Multiplexing
Mp4	✅	✅
Webm	✅	❌
Mov	✅	❌
Mkv	❌	❌
Avi	❌	❌

	Decoding	Encoding
Avc1	✅	✅
AAC	✅	✅ (Chromium only)
Opus	✅	✅
Wav	✅	✅
Hevc	✅	❌
VP9	✅	❌
VP8	✅	❌
Mp3	✅	❌
Ogg	✅	❌