Multi-lingual TTS
17 languages from one model
// official site: github.com ↗
AUDIO & MUSIC · PRO TIER
XTTS-v2 (Coqui)pro
XTTS-v2 is Coqui AI's multilingual text-to-speech model — 17 languages, voice cloning from 6-second samples, expressive emotional delivery, streaming output. Industry-leading open TTS, the canonical choice for self-hosted speech synthesis projects.
// What it is
A closer look.
XTTS-v2 is Coqui AI's multilingual text-to-speech model — 17 languages, voice cloning from 6-second samples, expressive emotional delivery, streaming output. Industry-leading open TTS, the canonical choice for self-hosted speech synthesis projects.
The voice equivalent of "open SDXL" — best-in-class open weights with permissive commercial terms.
// Use cases
What it's for.
Concrete scenarios where teams pick XTTS-v2 (Coqui) over the SaaS alternative.
Voice cloning
6-second sample → speech in cloned voice
Real-time streaming
chunked audio output, low latency
Cross-lingual generation
English speaker → speak in Spanish/French/Italian
Emotion-aware delivery
natural prosody, not robotic
API server
REST endpoints for programmatic use
// Who it's for
Built for these teams.
If your team profile matches one of these, XTTS-v2 (Coqui) is a strong fit out of the box.
Podcast producers
generating multi-language content
Game studios
creating character voices
Educational platforms
narrating content in multiple languages
Marketers
producing demo videos at scale
Accessibility teams
auto-narrating articles for screen readers
Hosting providers
selling voice synthesis services
// Differentiators
Why teams pick XTTS-v2 (Coqui).
When evaluating self-hosted options for this category, here are the dimensions on which XTTS-v2 (Coqui) consistently lands above the alternatives.
- ✓MPL-2.0 / CPML license — fully open; commercial OK with attribution
- ✓17 languages — broader coverage than F5-TTS, ChatTTS
- ✓Voice cloning quality — 6-second sample is impressive
- ✓Streaming server — production-ready API
- ✓Coqui pedigree — speech-tech veterans (formerly Mozilla DeepSpeech team)
- ✓Active community — frequent fine-tuned forks for specific languages
// Integrations
Connects to.
The stack you'll plug XTTS-v2 (Coqui) into — services, protocols, and adjacent apps in the BluixApps catalog.
◇
REST API server
/tts_stream endpoint for programmatic use
◈
WebSocket support
for real-time streaming
◆
Speaker library
reference voice samples stored persistently
▣
HuggingFace integration
model versions tracked
▦
Pair with Whisper
speech → text → translate → re-speak in new voice
▩
Pair with LLM
text generation → XTTS narration
// Adoption & deployment
Notable users & community
- 33k+ GitHub stars (parent Coqui TTS repo)
- Coqui AI (founded by ex-Mozilla DeepSpeech team)
- Industry-standard for open TTS
- Used in commercial products + academic research
- Active fine-tuning community on HuggingFace
What we ship
- Docker (ghcr.io/coqui-ai/xtts-streaming-server:latest-cuda121)
- Persistent volumes: models, output, speakers (reference voices)
- COQUI_TOS_AGREED=1 + MODEL_NAME pre-set
- Port 5002 (default XTTS) with Swagger docs at /docs
- Install report at
/root/bluixapps/xtts.txt - Acceptable Use Policy noted (no impersonation without consent)
- Sample API calls for voice cloning + text-to-speech in install report
- GPU pre-flight check via
bluixapps_ensure_nvidia_runtime - Backup hook covers speakers + outputs
// Tips & operations
Run it properly.
Operational guidance from running this in production — what to lock down, what surprises people.
// PERFORMANCE
Reference voice
6-30 seconds, clean speech, low noise, single speaker
// SECURITY
Languages supported
en, es, fr, de, it, pt, pl, tr, ru, nl, cs, ar, zh-cn, ja, hu, ko, hi
// OPERATIONS
VRAM
6 GB GPU optimal, 4 GB CPU fallback works
// RELIABILITY
Streaming mode
chunk size affects latency vs throughput tradeoff
// DEPLOYMENT
Speaker storage
/opt/xtts/speakers/ keeps your reference voices
// SCALING
Production
reverse proxy + auth, rate limiting via gateway
// MAINTENANCE
License caveat
voice cloning has misuse potential — disclose AI-generated audio
6144
// min ram (MB)
8
// min disk (GB)
5002
// access port
http
// protocol
pro
// bluixapps tier
// Alternatives in Audio & music
Compare with
Project resources
Official sitegithub.com ↗