XTTS v2 (Coqui)
Multilingual voice cloning in 6 seconds.
Open Source 4–6 GB VRAM
- Min VRAM
- 4 GB
- GPU class
- Entry GPU
- Quant
- FP16
Actually FreeNo SignupOpen SourceWatermark-Free
ElevenLabs
The benchmark commercial TTS / voice clone API.
Freemium · from $5/mo☁ Cloud · no GPUCloud API
Watermark-FreeHobbyist-FriendlyAPI
Visit ElevenLabsUpdated 2026-05-15 · Direct link
Screenshot placeholder · ElevenLabs
What is ElevenLabs?
ElevenLabs is the TTS service most production teams pick by default. Voice cloning is best-in-class, the latency is low enough for real-time agents, and Voice Design lets you describe a voice and get it. Closed-source and paid, but the quality bar everyone else is chasing.
Pros & cons
Pros
- The reference quality for English TTS and voice cloning
- Real-time TTS (<300ms first token) suitable for live agents
- Voice Design — synthesise a new voice from a text description
- Mature SDKs in 7+ languages
Cons
- Closed-source, usage-priced — costs add up at scale
- Voice cloning gated behind paid plans for safety/IP reasons
What's actually free?
10,000 chars/month free; paid plans from $5/mo (Starter) to $330/mo (Pro).
Watermark-Free
Alternatives
F5-TTS
Zero-shot voice cloning TTS — 15 s of audio is enough.
Open Source 8–12 GB VRAM
- Min VRAM
- 8 GB
- GPU class
- Mid GPU
- Quant
- FP16
Actually FreeNo SignupOpen SourceWatermark-Free
Bark
Suno's expressive transformer-based TTS.
Open Source 8–12 GB VRAM
- Min VRAM
- 8 GB
- GPU class
- Entry GPU
- Quant
- FP16
Actually FreeNo SignupOpen SourceWatermark-Free