Sopro V2: SOTA-level 120M param voice cloning TTS
SammyDaBeast · reddit · 2026-08-27
Sopro V2 is an open-source zero-shot voice cloning TTS with just 120M parameters, supporting English, Portuguese, French, and German. It is optimized for CPU inference.
Key Highlights:
- Small but Powerful: Achieves a WER of 1.51-1.65 on Seed-TTS-eval, outperforming models 3-14x larger (e.g., F5-TTS, CosyVoice).
- Low Latency Streaming: 300ms time-to-first-audio on M3 CPU, 0.07 RTF on H100.
- Accessibility: Includes a browser demo and Python package; runs locally with a single command.
- Language Support: The first open TTS to natively target European Portuguese (not Brazilian).
More from Multimodal
- Open-source LoRA turns MiniMax H3 into seamless 360-degree orbit videos from a single photo — camenduru · 2026-10-15
- Neon Noir Elegance prompt template goes viral with striking pink-blue two-tone outputs — LudovicCreator · 2026-10-15
- Krea 2 Turbo + px16 LoRA generates game-ready pixel art in 5s, where GPT Image 2.5 fails — Angaisb_ · 2026-10-15
- How Goldfish Made Its Kung Fu Scenes: Real Nanquan Choreography + Seedance 2.5 — JunieLauX · 2026-10-15
- Prompt share: 'Crimson Footsteps' makes one red element pop in monochrome AI photos — azed_ai · 2026-10-15
- Magnific One generates fake Zhangjiajie tourist photos indistinguishable from real shots — techhalla · 2026-10-15