跳到主要内容

Plugins

Plugins are available since Buzz version 1.4.5.

Plugins extend Buzz's transcription pipeline without requiring changes to the core application. They can process audio before transcription, modify transcription results, and run custom actions after a transcription is saved (such as exporting to additional formats or generating summaries).

To manage plugins:

  • Open Help → Plugins.
  • Enable or disable individual plugins using the toggle next to each one.
  • Drag plugins to change the order they run in.
  • Click the settings icon next to a plugin to configure its options.
  • Add a community plugin by clicking Add by URL and pasting a link to a .zip file.

Built-in plugins​

AI Summary​

Generates a summary of the transcript using an OpenAI-compatible API (e.g. OpenAI, Ollama) and saves it to the transcript's Notes field and/or a text file next to the source audio. Requires an API key and base URL.

Enhanced Language Detection​

Automatically detects the spoken language before transcription begins using a local Whisper model. Useful when transcribing files with unknown languages. Can optionally download the tiny Whisper model if no local model is available.

Export to DOCX​

Exports the completed transcript to a Microsoft Word .docx file. Timestamps can optionally be included. No extra dependencies required.

Resize Transcript​

Automatically regroups word-level transcript segments into subtitle-sized chunks after transcription. Supports merging by silence gaps, splitting on punctuation, and enforcing a maximum subtitle length. Requires word-level timings to be enabled on the transcription.

DeepFilterNet Noise Reduction​

Removes background noise from audio using the DeepFilterNet model before transcription begins. The denoised audio is saved as a new file with a _DeepFilterNet3.wav suffix next to the source file. By default this file is deleted after transcription completes; enable Keep denoised file after transcription in the plugin settings to retain it.

Skip Already Transcribed​

Skips transcription if results for the imported file have already been produced. Supports two detection methods (both can be enabled at the same time):

  • Check for existing result files (enabled by default) — looks for a .txt, .srt, or .vtt file matching the audio filename in the same folder. If found, the existing transcript is imported and the file is marked as Skipped without re-running the model.
  • Check in transcription database — checks whether this filename already has a completed transcription in Buzz's database. If found, the prior transcript segments are copied to the new entry and the file is marked as Skipped.

This is useful when re-importing a folder of files where some have already been transcribed, or when watching a folder automatically.

Building your own plugin​

The built-in plugins are a good starting point for writing a custom plugin. Browse the source code in buzz/plugins to see how they are structured and which hooks they use. Alternatively point your AI coding agent to this folder and ask to crete the plugin you need.

If you build something useful, please share it with the Buzz community! Open a GitHub Discussion or submit a pull request to have your plugin considered for inclusion as a built-in.