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
F5-TTS
Zero-shot voice cloning TTS — 15 s of audio is enough.
Open Source 8–12 GB VRAMRuns locally
Actually FreeNo SignupOpen SourceWatermark-FreeHobbyist-Friendly
Visit F5-TTSUpdated 2026-04-18 · Direct link
Hardware requirements
Runs locally · Mid GPU (12 GB)
Min VRAM
8 GB
Rec. VRAM
12 GB
Min RAM
16 GB
Rec. RAM
32 GB
Disk
15 GB
GPU class
Mid GPU
11.8+Apple Silicon ✓GPU RequiredQuant: FP16
8 GB VRAM viable with shorter contexts; 12 GB for full sequence lengths.
Screenshot placeholder · F5-TTS
What is F5-TTS?
F5-TTS is the current state-of-the-art open-weight zero-shot text-to-speech model. Give it a 15-second voice sample plus target text and it produces natural-sounding speech in that voice. Trained on 100k hours of multilingual audio; runs on a single 12 GB GPU.
Pros & cons
Pros
- Zero-shot cloning genuinely works on 15 s of clean audio
- Naturalness comparable to closed commercial TTS
- Active research lab maintenance (SWivid)
Cons
- Non-commercial license — not for paid products
- English / Chinese strongest; other languages weaker
What's actually free?
CC-BY-NC-4.0 — free for non-commercial use.
✓ Actually FreeNo SignupOpen SourceWatermark-Free
Alternatives
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
Tortoise TTS
Slow, but the quality is worth the wait.
Open Source 4–8 GB VRAM
- Min VRAM
- 4 GB
- GPU class
- Entry GPU
- Quant
- FP16
Actually FreeNo SignupOpen SourceWatermark-Free