Speech Model Match: a constraint-first guide to local ASR

Field notes

Users select language, hardware, workload, priority, and requirements including streaming, timestamps, and speaker diarization. The browser-based matcher compares those choices against a curated set of 18 model entries and returns a deterministic fit score rather than a universal accuracy ranking. Its comparison model keeps licenses separate, distinguishing MIT, Apache-2.0, CC BY 4.0, and custom terms, while linked primary sources provide a route for further checking.