Skip to content

Releases: joactr/meeting-assistant

Meeting Assistant 1.0.0-beta.6

Choose a tag to compare

@joactr joactr released this 26 Oct 07:40

Meeting Assistant 1.0.0-beta.6 for Windows x64.

Existing installs update in-app (signed updater payload). New installs: run the -setup.exe; it is not Authenticode-signed, so SmartScreen shows More info → Run anyway.

Runtime packs rebuilt in this release: pyannote. Reused from earlier releases: azure-speech, embeddings, phonon, sherpa, whisper, whisper-cuda.

Meeting Assistant 1.0.0-beta.5

Choose a tag to compare

@joactr joactr released this 24 Oct 22:05

Meeting Assistant 1.0.0-beta.5 for Windows x64.

Existing installs update in-app (signed updater payload). New installs: run the -setup.exe; it is not Authenticode-signed, so SmartScreen shows More info → Run anyway.

Runtime packs rebuilt in this release: none. Reused from earlier releases: azure-speech, embeddings, phonon, pyannote, sherpa, whisper, whisper-cuda.

Meeting Assistant 1.0.0-beta.4

Choose a tag to compare

@joactr joactr released this 24 Oct 20:28

Meeting Assistant 1.0.0-beta.4 for Windows x64.

Existing installs update in-app (signed updater payload). New installs: run the -setup.exe; it is not Authenticode-signed, so SmartScreen shows More info → Run anyway.

Runtime packs rebuilt in this release: none. Reused from earlier releases: azure-speech, embeddings, phonon, pyannote, sherpa, whisper, whisper-cuda.

Meeting Assistant 1.0.0-beta.3

Choose a tag to compare

@joactr joactr released this 24 Oct 19:06

Meeting Assistant 1.0.0-beta.3 for Windows x64.

Existing installs update in-app (signed updater payload). New installs: run the -setup.exe; it is not Authenticode-signed, so SmartScreen shows More info → Run anyway.

Runtime packs rebuilt in this release: none. Reused from earlier releases: azure-speech, embeddings, phonon, pyannote, sherpa, whisper, whisper-cuda.

Meeting Assistant 1.0.0-beta.2

Choose a tag to compare

@joactr joactr released this 24 Oct 08:37

Meeting Assistant 1.0.0-beta.2 for Windows x64.

Existing installs update in-app (signed updater payload). New installs: run the -setup.exe; it is not Authenticode-signed, so SmartScreen shows More info → Run anyway.

Runtime packs rebuilt in this release: none. Reused from earlier releases: azure-speech, embeddings, phonon, pyannote, sherpa, whisper, whisper-cuda.

Meeting Assistant 1.0.0-beta.1 (Windows, unsigned)

Choose a tag to compare

@joactr joactr released this 24 Oct 07:34

First public beta. Windows x64 only. Unsigned developer build.

Warning

The installer is not Authenticode-signed, so Windows SmartScreen will say "Windows protected your PC". Click More info → Run anyway only if you trust this repository. Verify downloads against SHA256SUMS.txt.

Install

  1. Download and run Meeting.Assistant_1.0.0-beta.1_x64-setup.exe.
  2. In the app, open Settings → Providers / models. Download a speech runtime and a model from inside the app. Each download is pinned by SHA-256, and runtimes are verified against an Ed25519-signed catalog embedded in the installer. You don't need to download the runtime zips below by hand; they're attached here because the app fetches them from this release.

Runtime packs (downloaded by the app on demand)

Pack Size What it's for
whisper 101 MB Default local transcription (CPU, INT8)
whisper-cuda 1.47 GB Optional NVIDIA GPU Whisper (driver 525+). Contains NVIDIA cuBLAS/cuDNN/NVRTC under NVIDIA's license, not MIT
phonon 261 MB Phonon-2.5, a fast 25-language model on CPU with word timings
sherpa 30 MB Parakeet TDT 0.6B v3 (ONNX)
pyannote 286 MB Speaker diarization (model is gated: accept pyannote's terms on Hugging Face)
embeddings 47 MB Local multilingual search
azure-speech 31 MB Azure Speech SDK (cloud, consent-gated)

Models are separate downloads from Hugging Face, each after you accept its license. See the README's model credits.

Known limitations

  • No auto-update in this beta. Install new versions manually.
  • macOS and Linux builds are not provided yet; CI only checks that the source builds there.
  • PDF/SRT export and Ask are preview features.
  • Transcripts, summaries and actions can be wrong. Review them before relying on them.