Subsai

Self-hosters running Whisper locally for subtitle generation

Subsai wraps Whisper and its faster variants into a usable subtitle generator with web UI, CLI, and Python package — the cleanest single project for running captioning locally without depending on cloud services. Active community, MIT-style license. Setup expects technical comfort.

Subsai is an open-source subtitle generation project that wraps OpenAI's Whisper and its community-maintained faster variants into a coherent toolkit with multiple interfaces. The web UI offers point-and-click transcription suitable for non-developers who can get past the initial install. The command-line interface enables scripting and batch processing. The Python package allows embedding subtitle generation into larger applications. For self-hosters, privacy-conscious teams, and developers who want to avoid sending audio to a third party, Subsai is one of the cleanest options in 2026. Faster-whisper, which is bundled as an option, runs roughly four times faster than the reference OpenAI implementation at comparable accuracy and lower memory usage. With a modern GPU, full-episode transcription completes in a fraction of real time. Without a GPU, expect transcription to run slower than real time, which limits practical throughput. Output covers SRT, ASS, SSA, SUB, JSON, TXT, and VTT formats. Subsai is not a caption stylist — there is no template library, no animated burn-in, no social-style highlight behaviour. The output drops cleanly into Aegisub or Subtitle Edit for further polish. Setup expects technical comfort: Python environments, dependencies, and optional GPU drivers. Documentation is community-maintained but generally clear. License is permissive, so embedding in commercial pipelines is straightforward.