stt

Why we included this project

This tool runs a speech-to-text pipeline locally on the faster-whisper models, so you can transcribe audio and video without sending media to a third-party service. Teams that care about privacy or per-minute API costs get the same practical accuracy as OpenAI's transcription API while keeping everything on their own hardware. A browser-based interface handles drag-and-drop uploads, lets you set the source language and pick a model from tiny up to large-v3, and returns results as plain text, JSON, or SRT subtitles with timestamps. The same engine is exposed through an HTTP endpoint, which makes it a straightforward substitute for an external speech recognition API in existing code. It is a good fit for generating subtitles for video content at scale or just cutting a recurring transcription bill.