Spaces:
Running
on
Zero
Running
on
Zero
new tts: MARS6
Browse files- app/models.py +28 -0
- test_tts_mars6.py +50 -0
app/models.py
CHANGED
@@ -78,6 +78,12 @@ AVAILABLE_MODELS = {
|
|
78 |
# llasa 3b TTS
|
79 |
'srinivasbilla/llasa-3b-tts': 'srinivasbilla/llasa-3b-tts',
|
80 |
|
|
|
|
|
|
|
|
|
|
|
|
|
81 |
# HF TTS w issues
|
82 |
# 'LeeSangHoon/HierSpeech_TTS': 'LeeSangHoon/HierSpeech_TTS', # irresponsive to exclamation marks # 4.29
|
83 |
# 'PolyAI/pheme': '/predict#0', # sleepy HF Space
|
@@ -363,6 +369,17 @@ HF_SPACES = {
|
|
363 |
'series': 'llasa 3b',
|
364 |
# 'emoji': '🥵', # requires 300s reserved ZeroGPU!
|
365 |
},
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
366 |
}
|
367 |
|
368 |
# for zero-shot TTS - voice sample used by XTTS (11 seconds)
|
@@ -569,6 +586,17 @@ OVERRIDE_INPUTS = {
|
|
569 |
'srinivasbilla/llasa-3b-tts': {
|
570 |
'sample_audio_path': handle_file('voice_samples/EN_B00004_S00051_W000213.mp3')
|
571 |
},
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
572 |
}
|
573 |
|
574 |
|
|
|
78 |
# llasa 3b TTS
|
79 |
'srinivasbilla/llasa-3b-tts': 'srinivasbilla/llasa-3b-tts',
|
80 |
|
81 |
+
# Mars5
|
82 |
+
# 'CAMB-AI/mars6-turbo-demo': 'CAMB-AI/mars6-turbo-demo',
|
83 |
+
|
84 |
+
# Mars6
|
85 |
+
# 'CAMB-AI/mars6-turbo-demo': 'CAMB-AI/mars6-turbo-demo',
|
86 |
+
|
87 |
# HF TTS w issues
|
88 |
# 'LeeSangHoon/HierSpeech_TTS': 'LeeSangHoon/HierSpeech_TTS', # irresponsive to exclamation marks # 4.29
|
89 |
# 'PolyAI/pheme': '/predict#0', # sleepy HF Space
|
|
|
369 |
'series': 'llasa 3b',
|
370 |
# 'emoji': '🥵', # requires 300s reserved ZeroGPU!
|
371 |
},
|
372 |
+
|
373 |
+
# Mars6
|
374 |
+
'CAMB-AI/mars6-turbo-demo': {
|
375 |
+
'name': 'MARS 6',
|
376 |
+
'function': '/inference',
|
377 |
+
'text_param_index': 'text',
|
378 |
+
'return_audio_index': 0,
|
379 |
+
'is_zero_gpu_space': False,
|
380 |
+
'is_closed_source': True,
|
381 |
+
'series': 'llasa 3b',
|
382 |
+
},
|
383 |
}
|
384 |
|
385 |
# for zero-shot TTS - voice sample used by XTTS (11 seconds)
|
|
|
586 |
'srinivasbilla/llasa-3b-tts': {
|
587 |
'sample_audio_path': handle_file('voice_samples/EN_B00004_S00051_W000213.mp3')
|
588 |
},
|
589 |
+
|
590 |
+
# MARS 6
|
591 |
+
'CAMB-AI/mars6-turbo-demo': {
|
592 |
+
'reference_audio': DEFAULT_VOICE_SAMPLE,
|
593 |
+
'reference_text': DEFAULT_VOICE_TRANSCRIPT,
|
594 |
+
'ras_K': 10,
|
595 |
+
'ras_t_r': 0.09,
|
596 |
+
'top_p': 0.2,
|
597 |
+
'quality_prefix': "48000",
|
598 |
+
'clone_method': "deep-clone",
|
599 |
+
},
|
600 |
}
|
601 |
|
602 |
|
test_tts_mars6.py
ADDED
@@ -0,0 +1,50 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
import os
|
2 |
+
from test_overrides import _get_param_examples, _override_params
|
3 |
+
from gradio_client import Client, file
|
4 |
+
|
5 |
+
model = "CAMB-AI/mars6-turbo-demo"
|
6 |
+
client = Client(model, hf_token=os.getenv('HF_TOKEN'))
|
7 |
+
endpoints = client.view_api(all_endpoints=True, print_info=False, return_format='dict')
|
8 |
+
# print(endpoints)
|
9 |
+
|
10 |
+
api_name = '/inference'
|
11 |
+
fn_index = None
|
12 |
+
end_parameters = None
|
13 |
+
text = 'This is what my voice sounds like.'
|
14 |
+
|
15 |
+
end_parameters = _get_param_examples(
|
16 |
+
endpoints['named_endpoints'][api_name]['parameters']
|
17 |
+
)
|
18 |
+
print(end_parameters)
|
19 |
+
|
20 |
+
|
21 |
+
space_inputs = end_parameters
|
22 |
+
# override some or all default parameters
|
23 |
+
space_inputs = _override_params(end_parameters, model)
|
24 |
+
|
25 |
+
if(type(space_inputs) == dict):
|
26 |
+
space_inputs['text'] = text
|
27 |
+
result = client.predict(
|
28 |
+
**space_inputs,
|
29 |
+
api_name=api_name,
|
30 |
+
fn_index=fn_index
|
31 |
+
)
|
32 |
+
else:
|
33 |
+
space_inputs[0] = text
|
34 |
+
result = client.predict(
|
35 |
+
*space_inputs,
|
36 |
+
api_name=api_name,
|
37 |
+
fn_index=fn_index
|
38 |
+
)
|
39 |
+
# space_inputs = {str(i): value for i, value in enumerate(space_inputs)}
|
40 |
+
|
41 |
+
print(space_inputs)
|
42 |
+
# print(*space_inputs)
|
43 |
+
# print(**space_inputs)
|
44 |
+
|
45 |
+
# result = client.predict(
|
46 |
+
# **space_inputs,
|
47 |
+
# api_name=api_name,
|
48 |
+
# fn_index=fn_index
|
49 |
+
# )
|
50 |
+
print(result)
|