Inflect-Nano-v1 — ONNX CPU Demo

Inflect-Nano-v1 — an ultra-small (4.63M params) English text-to-speech model, running here through its ONNX pipeline (encoder → numpy length-regulator → decoder → vocoder), entirely on CPU with onnxruntime. No GPU, no attention — a FastSpeech-style feed-forward acoustic model + a Snake-HiFiGAN vocoder, small and fast enough for on-device / edge inference.

Type English text below and click Generate. This is a research/demo model, not production-grade TTS.

Model weights: Luigi/Inflect-Nano-v1-ONNX · Original (PyTorch): owensong/Inflect-Nano-v1