AI Briefing#whisper
The latest AI and developer news about #whisper, with the original source and a short summary.
The latest AI and developer news about #whisper, with the original source and a short summary.
The Korean speech quality assessment model KOMOS has been released as open source, recording a higher correlation than UTMOS.
The method achieves a mean F1 score of 0.798 across 14 language pairs in CS-FLEURS using the Whisper-small encoder.
Both studies achieved an F1 score of 90.14% on the ADReSSo benchmark, surpassing the previous best record
The MCP server proofcut for Claude Code has released transcription-based automatic editing and self-verification rendering features.
Samsung Electronics has released an AI framework that analyzes lyrics and musical structure together to evaluate songs at a human-expert level.
Vellium v1.1.0 released an update supporting local STT/TTS and automatic llama.cpp detection.
Apple's new SpeechAnalyzer API overwhelmed on-device speech recognition performance, recording higher accuracy and faster speed than Whisper models.
The open-source audio-driven human video generation framework LongCat-Video-Avatar 1.5 has been released.
Whisper, an MCP server that provides real-time network infrastructure information to LLMs like Claude or Cursor, has launched.
Microsoft's VibeVoice transcribed 1 hour of audio in 8 minutes 45 seconds.
Box, a security-enhanced Android AI app that integrates LiteRT, llama.cpp, and more to operate fully offline, has been released.
It's a self-hosted video editor that combines local Whisper with subtitle editing.
An open weights model has been released that transcribes overlapping speech from multiple speakers using a single model.
For Somali TTS, XTTS V4 works best so far, but the LLM and Whisper layers remain bottlenecks.
A fully local voice assistant for Windows powered by Whisper and Ollama has been released.
Hugging Face has launched an Inference Endpoints option that boosts Whisper transcription performance by up to 8x using vLLM.
Morgan Stanley partnered with OpenAI to build AI solutions that boost financial advisors' work efficiency, achieving high adoption rates.
It introduces how to implement a custom handler that combines Whisper and Pyannote in Hugging Face Inference Endpoints to support speaker diarization and speculative decoding.
By applying Speculative Decoding, the inference speed of Whisper was doubled while maintaining accuracy.
Replit and AI Camp have launched a 4-hour course teaching how to use LLMs with Python.
This guide explains how to fine-tune the Whisper model on multilingual automatic speech recognition (ASR) datasets using Hugging Face Transformers.