Sopro V2 Turbo Open Source: 120M Voice Cloning Model, 5x Faster than Realtime
SammyDaBeast · reddit · 2026-08-30
Sopro V2 Turbo is an open-source 120M parameter TTS voice cloning model that runs 5x faster than realtime on laptop CPUs. Key features:
- Clones voice from 5-20s of audio.
- 300ms time to first audio on CPU.
- Supports English, European Portuguese, French, German.
- Available via Python API, local web UI, and a browser package using WebGPU/WASM.
More from Multimodal
- The math: 5 rerolls at 480p plus a 4K upscale costs $5 vs $9 at 720p — HeyAmit_ · 2026-10-15
- Dreamina workflow: generate at 480p, upscale to 4K, cut video costs 44% — HeyAmit_ · 2026-10-15
- One-prompt GPT-6 Astra writes two Palestrina-style motets in LilyPond — deanwball · 2026-10-15
- DiffSynth-Studio Adds OPD Workflow: Few-Step Diffusion Training in ~2 GPU-Hours — bdsqlsz · 2026-10-15
- AI film and TV hit the market: Prime Video picks up AI spin-off, pilots screen at Mipcom — lmoroney · 2026-10-15
- Ramayana film taps DNEG's massive VFX and AI-assisted production for global release — CurieuxExplorer · 2026-10-15