bnewton-genmedlabs commited on
Commit
a35c16e
·
verified ·
1 Parent(s): 3907f97

Upload README.md with huggingface_hub

Browse files
Files changed (1) hide show
  1. README.md +77 -0
README.md ADDED
@@ -0,0 +1,77 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ ---
2
+ language:
3
+ - en
4
+ - es
5
+ - fr
6
+ - de
7
+ - it
8
+ - pt
9
+ - pl
10
+ - tr
11
+ - ru
12
+ - nl
13
+ - cs
14
+ - ar
15
+ - zh
16
+ - ja
17
+ - ko
18
+ - hu
19
+ - hi
20
+ license: other
21
+ tags:
22
+ - text-to-speech
23
+ - tts
24
+ - xtts
25
+ - mobile
26
+ - pytorch
27
+ ---
28
+
29
+ # XTTS v2 Mobile Checkpoint
30
+
31
+ This repository contains the XTTS v2 model exported for mobile deployment.
32
+
33
+ ## Model Details
34
+
35
+ - **Model**: XTTS v2 (Coqui TTS)
36
+ - **Type**: Multilingual Text-to-Speech
37
+ - **Languages**: 17 languages supported
38
+ - **Sample Rate**: 24kHz
39
+ - **PyTorch Version**: 2.8.0
40
+
41
+ ## Files
42
+
43
+ - `xtts_v2_checkpoint.pth`: Full model checkpoint (1.78 GB)
44
+ - `xtts_v2_mobile.pth`: Mobile-optimized checkpoint (1.78 GB)
45
+ - `config.json`: Model configuration
46
+ - `manifest.json`: File manifest with SHA256 hashes
47
+
48
+ ## Usage
49
+
50
+ ### Android/iOS Integration
51
+
52
+ 1. Download the checkpoint file
53
+ 2. Load with LibTorch 2.8.x
54
+ 3. Implement tokenization on the app side
55
+ 4. Use the model for inference
56
+
57
+ ### Python Usage
58
+
59
+ ```python
60
+ import torch
61
+
62
+ # Load checkpoint
63
+ checkpoint = torch.load("xtts_v2_mobile.pth", map_location="cpu")
64
+ model_state = checkpoint["model_state_dict"]
65
+ config = checkpoint.get("config", dict())
66
+ ```
67
+
68
+ ## License
69
+
70
+ This model is subject to the Coqui Public Model License (CPML).
71
+ For commercial use, please contact: licensing@coqui.ai
72
+
73
+ ## Notes
74
+
75
+ - Exported from the official XTTS v2 model
76
+ - Requires text preprocessing on the application side
77
+ - Speaker embeddings should be computed separately