- Home
- Game Voice Generators
- Fish Speech
Fish Speech
What it is
Fish Speech is a free, open-source multilingual text-to-speech system for generating expressive audio. Its S2 model accepts natural-language tags for detailed emotion and prosody control, and it can clone a voice from a short reference sample, typically 10–30 seconds, without extra fine-tuning. Multi-speaker reference audio and speaker ID tokens support dialogue generated in one run. The repository describes S2 Pro as supporting more than 80 languages without phonemes or language-specific preprocessing. Users can run it through command-line tools, a WebUI, a server, or Docker, with documented CPU installation and GPU deployment. The installation guide lists Linux or WSL and 24 GB of GPU memory for inference. API access is available, and enterprise integrations include real-time pipelines, WebRTC, workflow automation, Vapi, Twilio, and Retell. The research license allows research and non-commercial use free of charge; commercial use requires a separate written license. The Free Tier costs nothing with 8,000 monthly credits and generation limits.
For and against
For
- Inline emotion and prosody controls
- Voice cloning from short references
- Supports multi-speaker dialogue
- More than 80 languages listed
- Command-line, WebUI, server, and Docker options
Against
- Commercial use requires a separate license
- Inference guide lists 24 GB GPU memory
- Linux or WSL is listed as required
- Free tier limits credits and generation
The verdict
Fish Speech offers expressive multilingual synthesis with cloning and multi-speaker generation. Check the hardware guidance and research-only license terms before planning commercial or production use.
Who it suits
Fish Speech suits developers and researchers needing multilingual, expressive speech generation, voice cloning, or self-hosted deployment. Commercial users need separate written licensing from Fish Audio.
Plans and prices
Free Tier
8,000 credits monthly · up to 7 minutes generation · up to 500 characters per generation · 3 public voice slots
Read 5 Oct 2026 · fish.audio
Plus
250,000 credits monthly · up to 200 minutes · up to 15,000 characters per generation · priority generation
Read 5 Oct 2026 · fish.audio
Pro
2,000,000 credits monthly · up to 1,620 minutes · 3 team seats · up to 30,000 characters per generation
Read 5 Oct 2026 · fish.audio
Max
25,000,000 credits monthly · up to 6,250 minutes · 10 team seats · 15 professional voice slots
Read 5 Oct 2026 · fish.audio
Enterprise
Pay as you go · zero data retention · on-premise deployment · SOC2 compliance
Read 5 Oct 2026 · fish.audio
On the record
- Commercial limits
- The model license does not grant commercial rights, which require a separate written license from Fish Audio.github.com · read 5 Oct 2026
- Emotion and prosody control
- S2 supports fine-grained inline control of prosody and emotion through natural-language tags.speech.fish.audio · read 5 Oct 2026
- Hardware requirement
- The installation guide lists 24 GB of GPU memory for inference and Linux or WSL as system requirements.speech.fish.audio · read 5 Oct 2026
- Integrations
- The enterprise page lists support for real-time pipelines, WebRTC infrastructure, workflow automation, Vapi, Twilio, and Retell.fish.audio · read 5 Oct 2026
- Languages
- The repository describes S2 Pro as supporting more than 80 languages without phonemes or language-specific preprocessing.github.com · read 5 Oct 2026
- License
- The Fish Audio Research License permits research and non-commercial use free of charge, while commercial use requires a separate written license.github.com · read 5 Oct 2026
- Multiple speakers
- S2 can use multi-speaker reference audio and speaker ID tokens to generate dialogue with multiple speakers in one generation.speech.fish.audio · read 5 Oct 2026
- Purpose
- Fish Speech is an open source multilingual text-to-speech system that generates expressive speech.github.com · read 5 Oct 2026
- Security and compliance
- Fish Audio says its SOC 2 Type II audit is underway and offers zero data retention, HIPAA-aligned configurations, and self-hosted deployments for enterprise workloads.fish.audio · read 5 Oct 2026
- Streaming
- The documentation describes SGLang serving optimizations including continuous batching, paged KV cache, CUDA graphs, and prefix caching.speech.fish.audio · read 5 Oct 2026
- Support
- Enterprise support includes 24/7 production support, a technical account manager, and a 99% uptime SLA.fish.audio · read 5 Oct 2026
- Voice cloning
- S2 supports voice cloning from short reference samples, typically 10–30 seconds, without additional fine-tuning.speech.fish.audio · read 5 Oct 2026
- Ways to run
- The installation guide documents command-line, WebUI, server, and Docker setups, including CPU installation and GPU deployment.speech.fish.audio · read 5 Oct 2026
