sim04ful
commited on
Commit
•
8a5d43b
1
Parent(s):
5e4143f
type casting
Browse files- arible_schema_power.json +78 -0
- handler.py +6 -6
arible_schema_power.json
ADDED
@@ -0,0 +1,78 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
{
|
2 |
+
"title": "AI Voice Cloner",
|
3 |
+
"description": "Clone a voice using AI",
|
4 |
+
"inputs": [
|
5 |
+
{
|
6 |
+
"name": "text",
|
7 |
+
"type": "text",
|
8 |
+
"description": "Text to be narrated",
|
9 |
+
"area": true,
|
10 |
+
"options": {
|
11 |
+
"min": 100,
|
12 |
+
"max": 50000
|
13 |
+
},
|
14 |
+
"title": "Content"
|
15 |
+
},
|
16 |
+
{
|
17 |
+
"name": "audio_urls",
|
18 |
+
"type": "constant",
|
19 |
+
"value": [
|
20 |
+
"https://pub-93685b189ac24b30839990a7d9a14391.r2.dev/attenborough_short.wav"
|
21 |
+
]
|
22 |
+
},
|
23 |
+
{
|
24 |
+
"name": "gpt_cond_len",
|
25 |
+
"type": "number",
|
26 |
+
"description": "Length of audio used for gpt latents.",
|
27 |
+
"title": "GPT Conditioning Length",
|
28 |
+
"options": {
|
29 |
+
"min": 6,
|
30 |
+
"max": 60
|
31 |
+
},
|
32 |
+
"slider_step": 0.5
|
33 |
+
},
|
34 |
+
{
|
35 |
+
"name": "gpt_cond_chunk_len",
|
36 |
+
"type": "number",
|
37 |
+
"description": "Length of audio chunks used for gpt latents.",
|
38 |
+
"title": "GPT Conditioning Chunk Length",
|
39 |
+
"options": {
|
40 |
+
"min": 6,
|
41 |
+
"max": 60
|
42 |
+
},
|
43 |
+
"slider_step": 0.5
|
44 |
+
},
|
45 |
+
{
|
46 |
+
"name": "max_ref_length",
|
47 |
+
"type": "constant",
|
48 |
+
"value": 30
|
49 |
+
},
|
50 |
+
{
|
51 |
+
"name": "temperature",
|
52 |
+
"type": "number",
|
53 |
+
"description": "Temperature for sampling.",
|
54 |
+
"title": "Temperature",
|
55 |
+
"options": {
|
56 |
+
"min": 0.0,
|
57 |
+
"max": 1.0
|
58 |
+
},
|
59 |
+
"slider_step": 0.1
|
60 |
+
},
|
61 |
+
{
|
62 |
+
"name": "repetition_penalty",
|
63 |
+
"type": "number",
|
64 |
+
"description": "Penalty for repetition.",
|
65 |
+
"title": "Repetition Penalty",
|
66 |
+
"options": {
|
67 |
+
"min": 1.0,
|
68 |
+
"max": 10.0
|
69 |
+
},
|
70 |
+
"slider_step": 0.1
|
71 |
+
},
|
72 |
+
{
|
73 |
+
"name": "language",
|
74 |
+
"type": "constant",
|
75 |
+
"value": "en"
|
76 |
+
}
|
77 |
+
]
|
78 |
+
}
|
handler.py
CHANGED
@@ -69,9 +69,9 @@ class EndpointHandler:
|
|
69 |
speaker_embedding,
|
70 |
) = self.model.get_conditioning_latents(
|
71 |
audio_path=audio_paths,
|
72 |
-
gpt_cond_len=model_input["gpt_cond_len"],
|
73 |
-
gpt_cond_chunk_len=model_input["gpt_cond_chunk_len"],
|
74 |
-
max_ref_length=model_input["max_ref_length"],
|
75 |
)
|
76 |
|
77 |
print("Generating audio")
|
@@ -81,10 +81,10 @@ class EndpointHandler:
|
|
81 |
text=model_input["text"],
|
82 |
speaker_embedding=speaker_embedding,
|
83 |
gpt_cond_latent=gpt_cond_latent,
|
84 |
-
temperature=model_input["temperature"],
|
85 |
-
repetition_penalty=model_input["repetition_penalty"],
|
86 |
language=model_input["language"],
|
87 |
-
enable_text_splitting=
|
88 |
)
|
89 |
audio_file = io.BytesIO()
|
90 |
torchaudio.save(
|
|
|
69 |
speaker_embedding,
|
70 |
) = self.model.get_conditioning_latents(
|
71 |
audio_path=audio_paths,
|
72 |
+
gpt_cond_len=int(model_input["gpt_cond_len"]),
|
73 |
+
gpt_cond_chunk_len=int(model_input["gpt_cond_chunk_len"]),
|
74 |
+
max_ref_length=int(model_input["max_ref_length"]),
|
75 |
)
|
76 |
|
77 |
print("Generating audio")
|
|
|
81 |
text=model_input["text"],
|
82 |
speaker_embedding=speaker_embedding,
|
83 |
gpt_cond_latent=gpt_cond_latent,
|
84 |
+
temperature=float(model_input["temperature"]),
|
85 |
+
repetition_penalty=float(model_input["repetition_penalty"]),
|
86 |
language=model_input["language"],
|
87 |
+
enable_text_splitting=False,
|
88 |
)
|
89 |
audio_file = io.BytesIO()
|
90 |
torchaudio.save(
|