Shypd.aiThe largest open index of AI toolstinydiarize is an Audio & Music tool that extends OpenAI's Whisper models to add speaker diarization. It labels who said what in a transcript, making it essential for conversation transcripts like meetings or podcasts.
Trending
Also listed in
About
tinydiarize is a minimal, interpretable extension of OpenAI's Whisper models designed to add speaker diarization with few extra dependencies. It uses a finetuned model that incorporates special tokens to mark speaker changes, leveraging both voice and semantic context to differentiate speakers. This approach offers a unique benefit compared to conventional methods. The tool provides a finetuned checkpoint for the `small.en-tdrz` model and example inference code. It also includes tools for comparison and analysis, such as a scoring tool to measure accuracy and a reference script for comparing diarization pipelines. Experimental support is available for `whisper.cpp`, allowing it to run on consumer hardware like MacBooks and iPhones with minimal code changes. While currently a prototype, it aims to provide a starting point for improving performance and extending support to multilingual and speech translation applications.
Capabilities
Pricing & Plans
Open Source
Free
FAQs
Trending
