Update app.py
Browse files
app.py
CHANGED
@@ -33,14 +33,8 @@ pipe_dict = {
|
|
33 |
}
|
34 |
|
35 |
title = """
|
36 |
-
|
37 |
-
|
38 |
-
Massively Multilingual Speech (MMS) models are light-weight, low-latency TTS models based on the [VITS architecture](https://huggingface.co/docs/transformers/model_doc/vits).
|
39 |
-
Meta's [MMS](https://arxiv.org/abs/2305.13516) project, aiming to provide speech technology across a diverse range of languages. You can find more details about the supported languages and their ISO 639-3 codes in the [MMS Language Coverage Overview](https://dl.fbaipublicfiles.com/mms/misc/language_coverage_mms.html),
|
40 |
-
and see all MMS-TTS checkpoints on the Hugging Face Hub: [facebook/mms-tts](https://huggingface.co/models?sort=trending&search=facebook%2Fmms-tts).
|
41 |
-
|
42 |
-
Coupled with the right data and the right training recipe, you can get an excellent finetuned version of every MMS checkpoints in **20 minutes** with as little as **80 to 150 samples**.
|
43 |
-
Training recipe available in this [github repository](https://github.com/ylacombe/finetune-hf-vits)!
|
44 |
"""
|
45 |
|
46 |
max_speakers = 1
|
@@ -129,47 +123,12 @@ with gr.Blocks(css=css) as demo_blocks:
|
|
129 |
with gr.Accordion("Datasets and models details", open=False):
|
130 |
gr.Markdown("""
|
131 |
|
132 |
-
For each language, we used 100 to 150 samples of a single speaker to finetune the model.
|
133 |
-
### Spanish
|
134 |
-
* **Model**: [Spanish MMS TTS](https://huggingface.co/facebook/mms-tts-spa).
|
135 |
-
* **Datasets**:
|
136 |
-
- [Chilean Spanish TTS dataset](https://huggingface.co/datasets/ylacombe/google-chilean-spanish).
|
137 |
-
### Tamil
|
138 |
-
* **Model**: [Tamil MMS TTS](https://huggingface.co/facebook/mms-tts-tam).
|
139 |
-
* **Datasets**:
|
140 |
-
- [Tamil TTS dataset](https://huggingface.co/datasets/ylacombe/google-tamil).
|
141 |
-
### Gujarati
|
142 |
-
* **Model**: [Gujarati MMS TTS](https://huggingface.co/facebook/mms-tts-guj).
|
143 |
-
* **Datasets**:
|
144 |
-
- [Gujarati TTS dataset](https://huggingface.co/datasets/ylacombe/google-gujarati).
|
145 |
### Marathi
|
146 |
* **Model**: [Marathi MMS TTS](https://huggingface.co/facebook/mms-tts-mar).
|
147 |
* **Datasets**:
|
148 |
-
- [Marathi TTS dataset](https://huggingface.co/datasets/ylacombe/google-chilean-marathi).
|
149 |
-
### English
|
150 |
-
* **Model**: [VITS-ljs](https://huggingface.co/kakao-enterprise/vits-ljs)
|
151 |
-
* **Dataset**: [British Isles Accent](https://huggingface.co/datasets/ylacombe/english_dialects). For each accent, we used 100 to 150 samples of a single speaker to finetune [VITS-ljs](https://huggingface.co/kakao-enterprise/vits-ljs).
|
152 |
-
|
153 |
""")
|
154 |
|
155 |
-
with gr.Accordion("Run VITS and MMS with transformers", open=False):
|
156 |
-
gr.Markdown(
|
157 |
-
"""
|
158 |
-
```bash
|
159 |
-
pip install transformers
|
160 |
-
```
|
161 |
-
```py
|
162 |
-
from transformers import pipeline
|
163 |
-
import scipy
|
164 |
-
pipe = pipeline("text-to-speech", model="kakao-enterprise/vits-ljs", device=0)
|
165 |
-
|
166 |
-
results = pipe("A cinematic shot of a baby racoon wearing an intricate italian priest robe")
|
167 |
-
# write to a wav file
|
168 |
-
scipy.io.wavfile.write("audio_vits.wav", rate=results["sampling_rate"], data=results["audio"].squeeze())
|
169 |
-
```
|
170 |
-
"""
|
171 |
-
)
|
172 |
-
|
173 |
|
174 |
language.change(lambda language: gr.Dropdown(
|
175 |
models_per_language[language],
|
|
|
33 |
}
|
34 |
|
35 |
title = """
|
36 |
+
Marathi Parkinson Enabler: Speaking is a big challenge during Parakinsons. Patients show slurred speech and cannot communicate effectively.
|
37 |
+
This is marathi text to speech model for parkinson users who want to communicate in Marathi.
|
|
|
|
|
|
|
|
|
|
|
|
|
38 |
"""
|
39 |
|
40 |
max_speakers = 1
|
|
|
123 |
with gr.Accordion("Datasets and models details", open=False):
|
124 |
gr.Markdown("""
|
125 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
126 |
### Marathi
|
127 |
* **Model**: [Marathi MMS TTS](https://huggingface.co/facebook/mms-tts-mar).
|
128 |
* **Datasets**:
|
129 |
+
- [Marathi TTS dataset](https://huggingface.co/datasets/ylacombe/google-chilean-marathi).
|
|
|
|
|
|
|
|
|
130 |
""")
|
131 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
132 |
|
133 |
language.change(lambda language: gr.Dropdown(
|
134 |
models_per_language[language],
|