RL10x commited on
Commit
4a43f91
·
verified ·
1 Parent(s): 14dba20

Model card: remove phoneme list, GGUF notes and limits

Browse files
Files changed (1) hide show
  1. README.md +0 -15
README.md CHANGED
@@ -50,13 +50,6 @@ re-implemented.
50
  | Synthesizer | band-limited pulse train at the predicted pitch + white noise, each filtered per frame (complex cepstrum, STFT overlap-add) |
51
  | Styles | neutral, conv, happy, sad, anger, fear, surprise, disgust |
52
 
53
- **Phonemes** (index = position): `<pad> <bos> <eos> <sp> <unk> , . ? ! ; : " ' ( ) — … k kʰ ɡ ɡʱ ŋ tʃ tʃʰ dʒ dʒʱ ɲ ʈ ʈʰ ɖ ɖʱ ɳ
54
- t̪ t̪ʰ d̪ d̪ʱ n p pʰ b bʱ m j ɾ l ʋ ʃ ʂ s ɦ ɽ ɽʱ q x ɣ z f ʒ ə aː ɪ iː ʊ uː eː ɛː oː ɔː æ ɔ`.
55
- Nasalized vowels (मैं, गाँव) use the vowel's symbol with the per-phoneme nasal flag set to 1.
56
-
57
- The GGUF files use the architecture name `tinyvoca`: hyperparameters are stored as `tinyvoca.*` keys, along with the
58
- style names, the symbol list and the phone/vowel id lists. Tensor names are the PyTorch state-dict names.
59
-
60
  ## Training
61
 
62
  - **Data:** [Rasa](https://huggingface.co/datasets/ai4bharat/Rasa) by AI4Bharat, Hindi, one female speaker, with
@@ -65,14 +58,6 @@ style names, the symbol list and the phone/vowel id lists. Tensor names are the
65
  - Losses: multi-resolution STFT + log-mel, plus duration, pitch and energy losses.
66
  - Waveform critics (multi-period and multi-resolution) from step 8k, at 0.7× the reconstruction gradient.
67
 
68
- ## Limits
69
-
70
- - **A faint mechanical buzz.** It is clearest on earbuds, and comes from a slightly over-smoothed (blurred) spectrum,
71
- the usual weakness of models trained to predict an average spectrum.
72
- - One voice only (female).
73
- - Reads Hindi in Devanagari script. Numbers, dates and symbols must be written out as Hindi words before phonemization.
74
- - Rare words are sometimes mispronounced.
75
-
76
  ## License and credits
77
 
78
  The weights are released under **CC BY 4.0**, following the training data:
 
50
  | Synthesizer | band-limited pulse train at the predicted pitch + white noise, each filtered per frame (complex cepstrum, STFT overlap-add) |
51
  | Styles | neutral, conv, happy, sad, anger, fear, surprise, disgust |
52
 
 
 
 
 
 
 
 
53
  ## Training
54
 
55
  - **Data:** [Rasa](https://huggingface.co/datasets/ai4bharat/Rasa) by AI4Bharat, Hindi, one female speaker, with
 
58
  - Losses: multi-resolution STFT + log-mel, plus duration, pitch and energy losses.
59
  - Waveform critics (multi-period and multi-resolution) from step 8k, at 0.7× the reconstruction gradient.
60
 
 
 
 
 
 
 
 
 
61
  ## License and credits
62
 
63
  The weights are released under **CC BY 4.0**, following the training data: