Inflect-Micro-v2: complete voice in 9.36M parameters
owensong / Inflect-Micro-v2 like 166 Text-to-Speech PyTorch English speech-synthesis local-tts cpu edge-ai small-model base-model vits 24khz License: apache-2.0 Model card Files Files and versions xet Community 2 Copy to bucket new Listen Evaluation 1. Fixed-voice English TTS with deterministic seeds, long-text handling, and CPU or CUDA inference. A note from Owen I built and funded Inflect v2 independently.
- ▪owensong / Inflect-Micro-v2 like 166 Text-to-Speech PyTorch English speech-synthesis local-tts cpu edge-ai small-model base-model vits 24khz License: apache-2.0 Model card Files Files and versions xet Community 2 Copy to bucket new Listen E
- ▪Fixed-voice English TTS with deterministic seeds, long-text handling, and CPU or CUDA inference.
- ▪A note from Owen I built and funded Inflect v2 independently.
Hacker News (Best) files mainly under programming. We currently carry 12 of its stories.
Opening excerpt (first ~120 words) tap to expand
owensong / Inflect-Micro-v2 like 166 Text-to-Speech PyTorch English speech-synthesis local-tts cpu edge-ai small-model base-model vits 24khz License: apache-2.0 Model card Files Files and versions xet Community 2 Copy to bucket new Listen Evaluation 1. Human blind preference 2. Predicted naturalness versus footprint 3. Intelligibility on unseen text 4. CPU runtime 5. Complete weight footprint Choose the right Inflect Run locally Install Python Download through the Hub ONNX Runtime Release profile Package map Limitations Responsible use License, integrity, and attribution Private training scope and contact Citation Inflect-Micro-v2 Complete local text-to-waveform speech synthesis under 10M parameters.
…
Excerpt limited to ~120 words for fair-use compliance. The full article is at Huggingface.
Discussion
0 commentsMore from Huggingface
