Starstruk Logo MarkStarstruk

0.2.4.dev0

QuentinFuxa/WhisperLiveKit0.2.4.dev0Aug 2, 2025by QuentinFuxa

AI Summary

This release focuses on bug fixes and core improvements for WhisperLiveKit, notably updating the SimulStreaming Whisper core from version 20230918 to 20250625 to fix tensor mismatch issues on some GPUs, and resolving a critical diarization queue bug that was sending the entire audio buffer instead of just the latest chunk.

Key Highlights

  • Updated SimulStreaming Whisper core from 20230918 to 20250625, fixing tensor mismatch on some GPUs due to triton version
  • Fixed critical diarization queue bug that was sending entire pcm_buffer instead of latest chunk
  • Improved punctuation splitting logic for better diarization results
  • Fixed warmup issues with >30s audio files in SimulStreaming
  • Updated Dockerfile with build-essential and newer PyTorch version

New Features

  • Enhanced use_punctuation_split logic for improved diarization results
  • Latest SimulStreaming version with fixes for long audio file warmup
  • Updated Docker environment with build-essential and PyTorch updates

Full Release Notes

## Bug Fixes

- **Diarization Queue Audio Overlap** Fixed a bug where `diarization_queue` was sent the entire `self.pcm_buffer` on every iteration, instead of just the latest chunk. [PR by @choomegan](https://github.com/choomegan) ([commit](https://github.com/QuentinFuxa/WhisperLiveKit/commit/64e44fb24f5eafedefb8df536877dc4da51f1352))

- **License Display Error** Fixed dual license warning display when using simulstreaming backend. https://github.com/QuentinFuxa/WhisperLiveKit/commit/46efbdf1d977e42a7c192d970f596ae97023dab5

## Enhancements

- **Improved Punctuation Splitting for Diarization** Enhanced the `use_punctuation_split` logic to improve diarization results. Commits:  [3ad3683](https://github.com/QuentinFuxa/WhisperLiveKit/commit/3ad3683ca7d5ab038f4be7f31f49d6ccfa955c26), [5b9977c](https://github.com/QuentinFuxa/WhisperLiveKit/commit/5b9977c9af66925fcf61490fa78a267df8312ca3), [56114d3](https://github.com/QuentinFuxa/WhisperLiveKit/commit/56114d3071a0bd8a73394a082afb7f694c9c67a1)

- **Deployment Guide Update** Fixed and clarified the Deployment Guide in the README. [PR by @luisla-rivas](https://github.com/luisla-rivas) ([commit](https://github.com/QuentinFuxa/WhisperLiveKit/commit/2564928d83e686800ea92693a26e8cca29de99cc))

- **Architecture Update** https://github.com/QuentinFuxa/WhisperLiveKit/commit/e40b5a3ea00a49bf93bd4be4d5ac2f69b3bd8c24

- **Dockerfile Improvements** Updated Dockerfile to install `build-essential` and update the PyTorch version.  - (Idea from [@callumgarven](https://github.com/callumgarven)) ([commit](https://github.com/QuentinFuxa/WhisperLiveKit/commit/e4140cd2991fcbd95ef15ac5a6f9c4a02d115e2d))


## Core Updates

- **Update to latest version of SimulStreaming** Fixes warmup with >30s audio files

- **SimulStreaming Whisper Core Update** Updated SimulStreaming whisper core from version 20230918 to 20250625. Solves [tensor mismatch on some gpus due to triton version](https://github.com/QuentinFuxa/WhisperLiveKit/issues/152)  - Commits:  [8e056cb](https://github.com/QuentinFuxa/WhisperLiveKit/commit/8e056cbdf2cfd7dcb08e5a73d66d51d41d21d679), [4cfed6e](https://github.com/QuentinFuxa/WhisperLiveKit/commit/4cfed6e98ee8156de512308eab9123ad48a133a8)