Client-side ASR uses compact models — faster-whisper compiled to WebAssembly, for example — to transcribe on the device. It protects privacy and cuts cost, at the price of fitting within limited local compute, often aided by WebGPU.
The book · Volume 6 of 7
AI for Video: A Complete Guide to Intelligence in Video Products: From Models and Inference to Computer Vision, Speech, Search, and Generative Video
Client-side ASR keeps audio on the device and removes per-minute bills, but it has to fit into the CPU and memory of a browser tab. Volume 6 covers speech in detail, ASR and diarization, with deployment topologies and model artifacts, so you can decide when client-side ASR is good enough and when to send audio to a server. Written by Nikolay Sapunov, CEO at Fora Soft.
Considering client-side ASR in the browser to cut transcription costs?
We will test client-side ASR on your target devices and audio and send a recommendation with accuracy, latency and the savings versus cloud. 250+ video products shipped since 2005.
