Reverb ASR+Diarization: Open-Source Long-Form Audio Tool

The developer behind Reverb ASR+Diarization released the project on Hacker News, claiming it is the best open-source ASR for long-form audio. The claim is unverified and based on the developer’s own benchmark.

Open-source ASR with integrated diarization reduces dependency on per-minute API pricing for transcription. For builders operating meeting bots, media archivists, or call-center analytics, this consolidates a two-stage pipeline (transcription plus speaker labeling) into a single self-hosted component. The immediate operational shift is a lower marginal cost for processing hours-long audio, which previously forced a choice between expensive vendor tiers or brittle multi-model orchestration.

Builders evaluating Reverb should test against their specific acoustic conditions, as long-form performance claims often degrade with overlapping speech or background noise. If accuracy holds, the workflow change is straightforward: replace cloud transcription webhooks with local batch processing for high-volume, non-real-time workloads. Second-order effect: as self-hosted diarization matures, pricing pressure on turnkey ASR APIs will increase, particularly for media and meeting transcription where volume is high and tolerance for latency is low.