CaptionFlow — on-device streaming ASR captions for OBS 0.2.0

Independent third-party plugin for OBS Studio that turns local audio into live captions on your machine. CaptionFlow is not developed by, endorsed by, or affiliated with the OBS Project. GitHub: https://github.com/XWHQSJ/captionflow Latest release: https://github.com/XWHQSJ/captionflow/releases/latest Report issues: https://github.com/XWHQSJ/captionflow/issues License: GPL-2.0-or-later Why this plugin? CaptionFlow keeps speech recognition local during use: your microphone or desktop audio is decoded by sherpa-onnx on your machine, and captions are written to a text file that OBS Studio can read. The first model download contacts the upstream model host; after that, captioning works offline with the cached model. The initial release focuses on low-latency English captions, a bilingual Chinese/English preset, and an optional delay-line mute filter for sensitive words. Features

• Real-time captions — low-latency partial results while speech is still in progress, written atomically to a text file any Text (GDI+ / FreeType 2) source can read.
• One-click model download — pick English / bilingual / tiny preset in the filter properties; the plugin downloads and extracts on demand, with optional SHA-256 verification.
• Sensitive-word mute — load a hotwords file (word :boost); the plugin delays output audio so it can retroactively beep out matches. Beep frequency and volume adapt to the speaker's F0 + RMS.
• Hardware providers — CPU (default), CUDA (Windows + NVIDIA), DirectML (Windows + any GPU).
• Supply-chain verifiable — every release ships with a Sigstore build provenance attestation, so you can verify the binary came out of public CI.
• Bilingual UI — English and Simplified Chinese (zh-CN) locale included.

Supported platforms

• Windows 10 / 11 x64 (OBS Studio 31.0+)
• macOS 11+ universal (Apple Silicon + Intel; OBS Studio 31.0+)

Linux users: the code compiles cleanly, builds aren't shipped yet — contributions welcome. Installation Download from the GitHub release linked above:

• Windows: extract the ZIP, merge obs-plugins\ and data\obs-plugins\ into %ProgramFiles%\obs-studio\.
• macOS: open the .pkg; it installs into the user OBS Studio plugin directory.

The current packages are unsigned. Download them only from the GitHub release page and verify the Sigstore build provenance attestation before installing: First use (60 seconds)

1. Right-click an audio source → Filters → + → CaptionFlow.
2. Click Download Model… and pick a preset.
3. Set Caption Output File to somewhere like /tmp/captions.txt.
4. Add a Text (GDI+) / Text (FreeType 2) source → enable Read from file → point at the same path.
5. Speak. Watch captions.

Model presets

• English (20M, fast) — ~70 MB — default streamers
• Chinese + English (bilingual) — ~300 MB — bilingual content
• English (tiny) — ~40 MB — low-end CPUs

License & AI tooling disclosure Licensed under GPL-2.0-or-later, matching OBS Studio. Full source code is published on GitHub. Per the OBS Forum resource policy (updated 2026-01-20): AI / LLM coding assistants were used during development for routine refactoring, boilerplate, and documentation drafting. All architectural decisions, the audio pipeline, the ASR integration, the muting algorithm, the unit tests, and every release artefact were authored, reviewed, and tested by a human maintainer. The plugin is not "vibe coded": all 45 offline unit tests pass under -Werror, AddressSanitizer and UndefinedBehaviorSanitizer in CI, and the code paths exercising sherpa-onnx were verified end-to-end on real OBS Studio installs. Support & issues Bug reports, feature requests, and general support: https://github.com/XWHQSJ/captionflow/issues Built on

• sherpa-onnx (Next-gen Kaldi team) — streaming ASR runtime
• obs-plugintemplate — build-system scaffolding
• OBS Studio plugin SDK