Fish Speech
Multilingual text-to-speech and voice cloning server (Fish Audio Drama 3 and S2-Pro) with an HTTP API and web UI.
Quick answer
Fish Speech is a multilingual text-to-speech and voice cloning server (Fish Audio Drama 3 and S2-Pro) with an HTTP API and web UI, listed under Speech & translation. A typical small-business use: generate narration or voice-overs for training videos and product demos in several languages. It needs a GPU server; we do not estimate that cost because GPU prices vary too much by card and provider. License: Fish Audio Community License (S2-Pro weights under Fish Audio Research License).
Last updated · facts verified against official docs on · sources
Activity data comes from GitHub and was last refreshed Nov 1, 2026.
What is Fish Speech?
Fish Speech is the open code behind Fish Audio's text-to-speech models. It runs the Drama 3 and S2-Pro models for multilingual speech generation, voice cloning from reference audio, multi-speaker scenes and inline emotion tags such as [whisper] or [laughing]. The Docker server image exposes an HTTP API and a built-in web UI on port 8080; a command-line tool is also included. Drama 3 adds local editing of generated audio and multi-turn revision.
Where it fits in a 10–200 person business
- Generate narration or voice-overs for training videos and product demos in several languages
- Give an internal assistant or kiosk a spoken voice without a cloud TTS API
- Clone a consenting staff member's voice for consistent brand audio
These use cases are our editorial suggestions, not claims made by the project.
Can a business use Fish Speech? (license)
| License | Fish Audio Community License (S2-Pro weights under Fish Audio Research License) |
|---|---|
| Type | Custom / conditional: Apache- or BSD-based with extra conditions (branding, multi-tenant SaaS, derivatives). |
| Internal business use | yes only while your organisation's annual revenue is under USD 500,000; above that a commercial license is needed |
| What to watch | Free for research and non-commercial use. Commercial use, including internal business operations, is allowed only while you and your affiliates have under USD 500,000 annual revenue; above that, or to offer a hosted TTS or voice-cloning service, you need a written license from Fish Audio. You must show "Built with Fish Audio", may not clone a real person's voice without consent, and may not use it or its output to train other speech models. The S2-Pro weights stay under the Fish Audio Research License. |
This is a plain-English summary of the license text, not legal advice. Third-party models you run with it have their own licenses.
Data & compliance: does Fish Speech phone home?
| Telemetry (usage data sent to the vendor) | Not documented |
|---|---|
| Other calls home by default (updates, marketplaces, default cloud services) | Not documented |
| Single sign-on (OIDC / SAML / LDAP) | Not documented |
| Audit log | Not documented |
| Role-based access control (RBAC) | Not documented |
| Data residency / GDPR notes from the project | Not documented |
Checked against the project's own docs and repository on . "Not documented" means we could not find the answer in official sources; it does not mean the feature is missing or that no data is sent. Third-party services you connect (model APIs, vector databases, search engines) have their own data policies. Compare all apps.
What hardware does Fish Speech need?
The project does not publish a minimum CPU/RAM figure, so we leave the official value blank rather than guess.
GPU: Required for Drama 3: NVIDIA GPU only, at least 48 GB GPU memory (FP8). S2-Pro needs at least 24 GB GPU memory and can also run on CPU (a CPU image exists for S2-Pro only).
No server CPU, RAM or disk figures are published; the requirements table lists GPU memory per model. Supported systems are Linux and WSL. AMD ROCm and Intel Arc are supported for S2-Pro only.
Source: speech.fish.audio/install/
How much does it cost to host Fish Speech per month?
Needs a GPU server for practical use. Not estimated, because GPU prices vary too much by card and provider.
How to deploy Fish Speech
Method: Docker (GPU)
pip install -U "huggingface_hub[cli]"
hf download fishaudio/s2-pro --local-dir checkpoints/s2-pro
docker run --gpus all \
-v ./checkpoints:/app/checkpoints \
-e LLAMA_CHECKPOINT_PATH=checkpoints/s2-pro \
-e DECODER_CHECKPOINT_PATH=checkpoints/s2-pro/codec.pth \
-p 8080:8080 \
fishaudio/fish-speech:server-cuda
# Web UI: http://localhost:8080/uiFrom the official instructions: speech.fish.audio/install/. Always check the source for the current version before you run it.
Known gotchas when self-hosting Fish Speech
- The license limits commercial use to organisations under USD 500,000 annual revenue and excludes hosted TTS or voice-cloning services for third parties; check it before any business use. source
- The API server only requires a bearer token when started with --api-key; the documented Docker commands do not set one, so keep the port off the public internet or add a key. source
- The containers expect model weights mounted at /app/checkpoints; download them before starting. Drama 3 falls back to the S2-Pro codec if its folder has no codec.pth, so download S2-Pro as well. source
Alternatives in Speech & translation
- whisper.cpp Fast C/C++ port of OpenAI Whisper speech recognition with a small HTTP transcription server.
- PDFMathTranslate PDF translator that keeps the original layout, formulas, charts and tables.
- Meetily Desktop meeting recorder that transcribes and summarizes meetings locally.
- LibreTranslate Self-hosted machine translation API with no dependency on Google or DeepL.
- Anarlog Local-first AI meeting notepad that records device audio without a bot joining the call.
- Speaches OpenAI-compatible speech-to-text and text-to-speech server (faster-whisper, Kokoro, Piper).
- Whisper ASR Webservice Whisper speech-to-text as a REST API in one Docker container.
- Scriberr Self-hosted web app that transcribes audio offline, with speaker labels and AI summaries.
Fish Speech FAQ
How much RAM does Fish Speech need?
Fish Speech does not publish an official RAM figure, so we do not state one. No server CPU, RAM or disk figures are published; the requirements table lists GPU memory per model. Supported systems are Linux and WSL. AMD ROCm and Intel Arc are supported for S2-Pro only.
Does Fish Speech need a GPU?
For practical use, yes. Required for Drama 3: NVIDIA GPU only, at least 48 GB GPU memory (FP8). S2-Pro needs at least 24 GB GPU memory and can also run on CPU (a CPU image exists for S2-Pro only).
How much does it cost to self-host Fish Speech?
The software is free to download. It needs a GPU server; we do not estimate that cost because GPU prices vary too much by card and provider.
Is Fish Speech free for commercial use?
Fish Speech is licensed under Fish Audio Community License (S2-Pro weights under Fish Audio Research License) (Custom / conditional). yes only while your organisation's annual revenue is under USD 500,000; above that a commercial license is needed Free for research and non-commercial use. Commercial use, including internal business operations, is allowed only while you and your affiliates have under USD 500,000 annual revenue; above that, or to offer a hosted TTS or voice-cloning service, you need a written license from Fish Audio. You must show "Built with Fish Audio", may not clone a real person's voice without consent, and may not use it or its output to train other speech models. The S2-Pro weights stay under the Fish Audio Research License. This is a summary, not legal advice.
What are common problems when self-hosting Fish Speech?
The license limits commercial use to organisations under USD 500,000 annual revenue and excludes hosted TTS or voice-cloning services for third parties; check it before any business use. The API server only requires a bearer token when started with --api-key; the documented Docker commands do not set one, so keep the port off the public internet or add a key. The containers expect model weights mounted at /app/checkpoints; download them before starting. Drama 3 falls back to the S2-Pro codec if its folder has no codec.pth, so download S2-Pro as well.
Is Fish Speech still maintained?
Yes, it is actively developed. Last commit: Oct 29, 2026. Latest release: v1.5.1 (May 31, 2025). Checked Nov 1, 2026.
How do I install Fish Speech?
The official method is Docker (GPU). The exact commands are in the "How to deploy" section above and in the official instructions: https://speech.fish.audio/install/
What are self-hosted alternatives to Fish Speech?
Other apps in Speech & translation: whisper.cpp, PDFMathTranslate, Meetily, LibreTranslate, Anarlog, Speaches, Whisper ASR Webservice, Scriberr.
Sources
- GitHub repository: github.com/fishaudio/fish-speech
- Official documentation, Hardware requirements, Deploy instructions, Gotcha 3: speech.fish.audio/install/
- License text, Gotcha 1: github.com/fishaudio/fish-speech/blob/main/LICENSE
- Gotcha 2: speech.fish.audio/inference/
- Hetzner Cloud pricing (cost estimate): www.hetzner.com/cloud/regular-performance
- GitHub stars, commits and releases: GitHub REST API, fetched
Facts on this page were checked against the project's own repository and docs on 2026-11-01. Spotted something out of date? Tell us.