Video Editing in the Browser
summary
Browsers now expose web codecs and ML libs like tensorflow.js and whisper so encoding, background removal and speech-to-text are possible, but practical limits matter: a 1000×1000 frame is ~4MB, 60 FPS becomes ~200MB/s before compression and common techniques give ~10x reduction; frames use key/delta/B‑frames so decode order differs from presentation order and decoding one packet can produce zero or many frames; copying into wasm costs negate gains so wasm is not faster; a proof-of-concept can decode‑reencode works as fast as Final Cut Pro but it needs hundreds of lines of brittle glue.
AI-written summary · about
chapters (12)
- 0:00Intro
- 1:06What is video editing
- 3:05Making a video
- 4:10Video size
- 5:10Image compression
- 6:56Continuous functions
- 9:45Hoffman encoding
- 10:33Video correct
- 12:06Bidirectional frames
- 14:36Decoder API
- 16:46Demoxing
- 21:59Conclusion
related talks
27mVideo Editing in the Browser
12mExploring Video Editing in the Browser
31mSplit Rendering TV Applications Using WASM
26mOptimizing Video Performance for Web Developers
47mVideo transcoding on the web
12mEditing video simply by editing text, in the browser: a meta story
10mVideo vectorization
21mDynamic video in the browser