mashreghi
Re:Professional video editing, right in the browser with WebGPU and WASM
on Sunday March 22, 2026 @04:36AM (#47475565)
"Professional" = people getting paid, not feature checklists.
Re:Professional video editing, right in the browser with WebGPU and WASM
on Sunday March 22, 2026 @04:35AM (#47475558)
If your baseline is Resolve, sure. But most people aren't cutting Hollywood timelines.
Re:Professional video editing, right in the browser with WebGPU and WASM
on Sunday March 22, 2026 @04:34AM (#47475551)
Different use case. "Runs everywhere instantly" beats "installs + config" for a lot of workflows.
Re:Professional video editing, right in the browser with WebGPU and WASM
on Sunday March 22, 2026 @04:00AM (#47475391)
Sure, and most social/video workflows fit comfortably under that. Not everything is 8K RAW timelines.
Re:Professional video editing, right in the browser with WebGPU and WASM
on Sunday March 22, 2026 @03:59AM (#47475390)
Most real-world edits aren't hitting those limits. Constraint ≠ useless, it just defines the target use case.
Re:Professional video editing, right in the browser with WebGPU and WASM
on Sunday March 22, 2026 @03:52AM (#47475351)
If people can ship paid work with it, it's professional enough.
Re:Professional video editing, right in the browser with WebGPU and WASM
on Sunday March 22, 2026 @03:50AM (#47475344)
Same argument was made about Figma vs native tools, didn't age well.
Re:Professional video editing, right in the browser with WebGPU and WASM
on Sunday March 22, 2026 @03:49AM (#47475340)
Photopea wasn't "Photoshop replacement" either, still massively useful.
Re:Subformer: Multilingual video dubbing with speaker diarization and ...
on Sunday January 11, 2026 @10:16PM (#46583632)
Hi HN,
We built Subformer (https://subformer.com), a web app that dubs videos into other languages while keeping speaker identity intact.
Most “AI dubbing” pipelines are just ASR → translation → TTS, which breaks as soon as you have multiple speakers. We instead run:
- VAD + speaker diarization - Audio Demixing - Global speaker clustering - Per-segment ASR + translation - Per-speaker TTS (voice cloning or synthetic) - Timeline-aligned remuxing back into the video
The tricky parts were diarization drift on long videos, timing mismatches after translation, and keeping costs sane when doing multilingual TTS at scale.
It’s still early, but it already works well for things like interviews, TV clips, and YouTube videos with multiple speakers.
Would love feedback from people who work on audio, speech, or localization.