Upload README.md with huggingface_hub

Upload README.md with huggingface_hub

Upload README.md with huggingface_hub

Card: fp16@2thr as CPU default; remove q8 row; honest int8 verdict (unfit on ARMv8.0 CPU)

Publish v2 deployment variant / card update: README.md

Model card: document PrimeTTS v2 (MB-iSTFT-VITS 34.7M, Xinran, 16 kHz) as flagship; v1 kept as legacy CPU family

Card: fix stale 16k-roadmap note (now shipped)

Card: promote 16 kHz flagship (v1b_16k, snake_16k, PESQ 2.70); 8 kHz kept as leanest on-device

Model card: rewrite for current on-device v1b (8k, 4.09M, snake_8k_lite) + vocoder-family/param table; synthesized + shortened (424->155 lines)

3-pass diagnostic: naturalize utterances + add per-model raw transcripts vs ground truth (errors highlighted); correct the P2 English OOD finding

Add 3-pass phonemic diagnostic (zh 38/38, en 39/39, entities) + teacher/v1/v2 benchmark (X-ASR+Breeze) + diagnosis

Add pronunciation diagnostic: X-ASR vs Breeze-ASR-25 cross-check (teacher/v1-4.63M/v2-6.85M) on the entity sentence

README: full v4 architecture/training/dataset + reproduction; v3 kept in v3_4.6M/

README: v4-50k now default (6.85M, clarity+timing); v3 kept for rollback

card: v3 pronunciation fix (ㄜ/ㄟ/ㄡ + ㄭ), zh-CER .106->.087

Card: ship 60k (clarity PESQ 3.22 + zh-CER .106), update metrics

Model card: 24kHz CC0 release, entity-coverage corpus (29.5k), updated metrics

README: reference voice = Mozilla Common Voice zh-TW (CC0/public-domain, commercial-clear), replacing Edge-TTS

README: one-command train-on-your-own-voice section

README: document VoxCPM2-TW teacher, v1 warm-start, ASR gate, GAN recipe + real metrics/data

Fix demo link: PrimeTTS-app (deleted) -> PrimeTTS-vs-Inflect-Nano-v1

demo URL -> Luigi/PrimeTTS-app

README: clearer layout + step-by-step training guide + quickstart

base_model_relation: finetune (not quantized)

weights at root; point demo link to Luigi/PrimeTTS-demo

PrimeTTS: full training pipeline + weights (fine-tune of Inflect-Nano-v1)