Speech & translation

Transcription, text-to-speech and machine translation you can run in-house.

Quick answer

We track 2 self-hostable speech & translation apps: LibreTranslate, Speaches. Most GitHub stars: LibreTranslate (17.1k). None of them publishes an official CPU/RAM figure we could use for a server-cost estimate.

Last updated Oct 22, 2026

Speech & translation: side-by-side

App	License	Official hardware	Est. server cost	GitHub stars	Status
LibreTranslate	AGPL-3.0 (Network copyleft (AGPL))	Not published	Not estimated (no official RAM figure)	17.1k	Active
Speaches	MIT (Permissive)	Not published	Not estimated (no official RAM figure)	3.7k	Stale (no commit in 180+ days)

All speech & translation apps

LibreTranslate

Active

Self-hosted machine translation API with no dependency on Google or DeepL.

★ 17.1k
⟳ Oct 12, 2026
Network copyleft (AGPL)
RAM: n/a
Cost: n/a

Speech & translation

Speaches

Stale

OpenAI-compatible speech-to-text and text-to-speech server (faster-whisper, Kokoro, Piper).

★ 3.7k
⟳ Apr 18, 2026
Permissive
RAM: n/a
Cost: n/a

Speech & translation