Speaches

Speaches is an OpenAI API-compatible server for self-hosted speech transcription, translation, and text-to-speech generation.

  • AI
  • Other
  • Productivity
  • Tools & Utilities
Speaches is an OpenAI API-compatible server for self-hosted speech transcription, translation, and text-to-speech generation. It uses faster-whisper for speech-to-text and piper or Kokoro for text-to-speech, so existing OpenAI-compatible tools can run local speech models on a home server.

Platform availability

Available on 1 of 10 platforms

  • Unraid
  • TrueNAS (not listed)
  • Umbrel (not listed)
  • ZimaOS (not listed)
  • Proxmox (not listed)
  • Helm (not listed)
  • Mac (not listed)
  • Windows (not listed)
  • Linux (not listed)
  • Any server (not listed)

Health score?A 0-100 score computed nightly from four factors: maintained (40%: recent commits and steady activity), popular (25%: GitHub stars, log scale), easy to install (20%: platforms, install notes, setup guides, Docker image) and light to run (15%: minimum RAM, ARM support). When the source code has been checked, a fifth factor, security check (20%: automated scan for committed secrets, unsafe desktop-app settings, known-vulnerable dependencies and project hygiene; any serious finding caps it at 20), is added. Factors without data are left out and the rest are rescaled.

59/100 Fair

Maintained
51last commit 217 days ago
Popular
723,775 GitHub stars
Easy to install
351 install platform, install notes, Docker image
Light to run
No data
Security check
82no serious findings, 3 minor, CI, tests

Based on 4 of 5 factors.

Availability: platform listings confirmed today, links checked 18 days ago

GitHub stars
3,775
Open issues
178
Last commit
2026-04-18
Latest release
v0.9.0-rc.3
License
MIT
Activity
stale

Checked today - source: GitHub (speaches-ai/speaches)

Resources & Compatibility

Requirements not published yet.

Checked 54 days ago - source: Docker Hub image tags and published docs

First-install notes

  • Run the CPU image on port 8000 Run the CPU Docker image in the background, with port 8000 published and a named Docker volume that holds downloaded model files. If port 8000 is blocked, other apps can’t reach Speaches, and if the model cache volume is missing, downloaded models may disappear when the container is removed. source
  • Set SPEACHES_BASE_URL before using the CLI Set SPEACHES_BASE_URL to your Speaches server address, such as http://localhost:8000, before running speaches-cli commands. Without it, the CLI does not know where to send requests, so model downloads and lists fail. source
  • Download a speech model after startup After the server is running, download at least one model with the CLI so voice features have something to load. Without a downloaded model, speech requests have no model to use and will fail. source
  • NVIDIA GPU use on Unraid For NVIDIA GPU use on Unraid, install the NVIDIA drivers and add --runtime=nvidia --gpus=all to extra parameters in advanced view. source

Fetched 48 days ago - each note links to its source

Alternatives

No alternatives collected yet.

Common questions

Which Docker Compose file should I use for Speaches?

Use the provided Docker Compose files: compose.cpu.yaml for CPU, compose.cuda.yaml for CUDA, and compose.cuda-cdi.yaml when CDI is enabled. Start the stack with docker compose up --detach.

Do I need extra steps before using text-to-speech?

Yes, the installation documentation warns that additional steps are required to use the text-to-speech feature. You should follow the Text-to-Speech documentation before expecting TTS to work.

How do I enable NVIDIA GPU support in Unraid?

Install the NVIDIA drivers and add --runtime=nvidia --gpus=all to the extra parameters in advanced view.

Answers sourced from ca.unraid.net, speaches.ai

Community

Collected 49 days ago - public community threads