ffmpeg-webCLI v2.0.0

Browser-based video editing, fully on-device. No uploads, no servers, no install -- your files never leave your device. This release turns ffmpeg-webCLI from a single-operation converter into a complete short-form video workflow, with on-device auto-captioning as the centerpiece.

Highlights

• Auto-Caption (Whisper, on-device). Generate captions from speech without uploading a thing. Audio is transcribed locally via Whisper (Transformers.js), the transcript is fully editable before you commit, and captions are embedded as a soft subtitle track or burned directly into the video. The entire pipeline: extract, transcribe, edit, embed, all runs in your browser. Your audio is never sent anywhere.
• Hardburned captions. Burn captions directly into the picture for platforms that require them (TikTok, Reels, Shorts). Rendered via the Canvas API with anti-aliased text, automatic word wrap, and configurable positioning. Ideal for muted-autoplay short-form video.
• Operation chaining (Stack mode). Stack multiple compatible operations and run them in a single pass, encoded only once, no generational quality loss from repeated re-encoding. A live composed-command preview shows the exact ffmpeg invocation before you run it.
• Batch processing. Apply an operation or an entire operation chain to many files at once. Per-file progress and preview, individual downloads, and ZIP-all export. Chain + batch combine: build a chain once, apply it to every file.

Also new

30+ operations, 25+ presets spanning conversion, editing, audio, compositing, effects, and social formats
Pad / Letterbox with aspect-ratio presets (16:9, 9:16, 1:1, 4:3, 4:5, 21:9) for every platform
Audio normalization (EBU R128 loudnorm) with YouTube/Spotify, podcast, and broadcast targets
Mix audio, picture-in-picture, side-by-side, and concatenate multi-input operations
Editable transcript workflow for captions, with SRT timing preserved
Expanded example command library (13 one-click recipes) and raw ffmpeg mode with live command preview

Reliability & platform

• Offline-first PWA -- install to your home screen; after first load it works with no network at all
• Screen Wake Lock keeps the screen active during long encodes on mobile and desktop
• Service worker correctly preserves COOP/COEP isolation headers for SharedArrayBuffer in offline mode
• Web Worker isolation keeps the UI responsive during processing
• Failed operations clean up the virtual filesystem automatically

Privacy

Zero egress, verifiable. Open DevTools → Network and watch: no outbound requests during processing. Only the ffmpeg-core WebAssembly binary and Whisper model weights are downloaded — once — and cached locally. Nothing else leaves your device, ever.

Good to know

Best for short-to-moderate clips. WebAssembly encoding is software-only (no hardware acceleration), so processing is slower than native ffmpeg -- fine for short clips, slower for long-form content. ffmpeg-webCLI is built for ad-hoc edits and short-form video, not bulk transcoding.
Hardburned captions cover common burned-in styles but do not replicate advanced libass effects (karaoke, complex ASS positioning).
Memory ceiling ~2 GB (WebAssembly limit). Very large files (>2 GB) or whole-clip reversal of long videos may exceed it.

Thanks

Operation chaining and batch processing were built in response to community feature requests. Thanks to everyone who filed issues, opened PRs, and tested — this release is shaped by your feedback.
Try it: https://tejaswigowda.com/ffmpeg-webCLI/