WhisperLiveKit Releases

A major release expanding the LLM ecosystem with four new Chinese providers, introducing analytics dashboards, and adding runtime flexibility.

Apr 11, 2026

Adds installation support for Debian-based Linux distributions and includes development work.

Apr 9, 2026

Adds AWS Bedrock provider, new search providers, and email login flow, while fixing numerous bugs including OAuth deprecation and UI issues.

Apr 9, 2026

This release adds 76 new templates and 42 CVEs, focusing on critical vulnerabilities such as authentication bypasses, remote code execution, and information disclosures across various software. It includes significant bug fixes for template execution errors and improvements to reduce false positives in detection templates.

Mar 31, 2026

Introduces a dedicated 'Security Analysis Mode' for the cr-agent, utilizing specialized LLM prompts for sink localization and post-filtering to isolate exploitable vulnerabilities. Includes fixes for deduplication logic, context assembly, and CLI flag handling.

Mar 25, 2026

Introduces 94 new templates and 47 CVEs, highlighting critical exploits like unauthenticated RCEs in Qwik and Langflow, and authentication bypasses in Ivanti and SolarWinds products. Also resolves bugs related to UniFi templates and false positives in MongoDB and JetBrains TeamCity detection.

Mar 16, 2026

Focuses on file management enhancements for Cloudreve, adding features for directory direct links and global default application settings. Fixes bugs related to OAuth client secret generation, pagination crashes, and Windows client file list inconsistencies.

Mar 14, 2026

Introduces Qwen3-ASR Beta backend with ForcedAligner word-level timestamps and auto language detection, adds DiffTracker for bandwidth-efficient WebSocket communication, and provides thread-safe per-session language override.

Mar 12, 2026

A major upgrade introducing support for advanced reasoning models, intelligent token caching, REST API access for analytics, and Langfuse v3 observability.

Feb 25, 2026

This release introduces the Voxtral backend with two variants: voxtral-mlx for native Apple Silicon performance (0.18-0.32x real-time) and voxtral (HF) for HuggingFace transformers on Linux/GPU. It also adds a comprehensive offline benchmark harness for testing all backends with WER, RTF, and timestamp accuracy metrics.

Feb 23, 2026

Updates PDF parsing capabilities to improve table extraction, support wide tables, and handle partially numbered lists.

Feb 20, 2026

A minor patch fixing document search links and preventing crashes during file transfers in slave nodes.

Feb 15, 2026

Bug fix release addressing critical SimulStreaming VRAM leaks, translation functionality issues, and various type errors. Also adds probability field support to ASRToken and improves DiartDiarization interface.

Feb 13, 2026

A patch release addressing Windows path issues in the exe launcher, adding alerts for spell check errors, and standardizing HTML manuscript formatting.

Jan 25, 2026

Focuses on resolving LiteLLM passthrough issues, fixing Windows file path compatibility, and improving Ollama configuration and installer stability.

Jan 17, 2026

Minor improvement release focusing on KV cache cleanup fixes, potential thread conflict resolution, and refined diarization timing.

Jan 16, 2026

This release focuses on stability and reliability by improving error diagnostics, enhancing DuckDuckGo search functionality, and adjusting OpenAI model configurations. It also introduces guardrails bypass capabilities and various prompt refinements.

Jan 6, 2026

Introduces native MLX AlignAtt for faster decoding and implements one shared ONNX Silero VAD model instance across sessions for better multi-speaker support.

Dec 9, 2025

Significant scalability improvements are achieved by sharing the Whisper decoder across multiple sessions, removing a previous bottleneck.

Nov 25, 2025

Performance improvements for VAD cuts and audio buffering, plus core evolutions including LoRa extension loader and CoreML encoder compatibility.

Nov 21, 2025