awesome-diarization

Why we included this project

Speaker diarization, the task of figuring out who spoke when in an audio recording, is a niche problem where good starting points are hard to find. This curated index is the most complete single entry point we know of, gathering survey and research papers, open-source frameworks, evaluation tools, clustering and speaker-embedding libraries, plus the datasets and noise sources you would need to train or benchmark a system. The organization helps you skip the digging: sections are split by task stage, from embedding and change detection to clustering and evaluation, and by research direction such as online diarization, joint ASR, audio-visual work, and LLM-based post-processing. For a team deciding whether to build diarization in-house or pick an existing library, this list maps the current state of the field and saves days of scattered searching.