model card and meta info
Browse files
README.md
CHANGED
@@ -1,5 +1,79 @@
|
|
1 |
---
|
2 |
license: mit
|
3 |
---
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
4 |
|
5 |
-
# Amphion Singing Voice Conversion Pretrained Models
|
|
|
1 |
---
|
2 |
license: mit
|
3 |
---
|
4 |
+
# Amphion Singing Voice Conversion Pretrained Models
|
5 |
+
|
6 |
+
## Quick Start
|
7 |
+
|
8 |
+
We provide a [DiffWaveNetSVC](https://github.com/open-mmlab/Amphion/tree/main/egs/svc/MultipleContentsSVC) pretrained checkpoint for you to play. Specially, it is trained under the real-world vocalist data (total duration: 6.16 hours), including the following 15 professional singers:
|
9 |
+
|
10 |
+
| Singer | Language | Training Duration (mins) |
|
11 |
+
| :-----------------: | :------: | :----------------------: |
|
12 |
+
| David Tao 陶喆 | Chinese | 45.51 |
|
13 |
+
| Eason Chan 陈奕迅 | Chinese | 43.36 |
|
14 |
+
| Feng Wang 汪峰 | Chinese | 41.08 |
|
15 |
+
| Jian Li 李健 | Chinese | 38.90 |
|
16 |
+
| John Mayer | English | 30.83 |
|
17 |
+
| Adele | English | 27.23 |
|
18 |
+
| Ying Na 那英 | Chinese | 27.02 |
|
19 |
+
| Yijie Shi 石倚洁 | Chinese | 24.93 |
|
20 |
+
| Jacky Cheung 张学友 | Chinese | 18.31 |
|
21 |
+
| Taylor Swift | English | 18.31 |
|
22 |
+
| Faye Wong 王菲 | English | 16.78 |
|
23 |
+
| Michael Jackson | English | 15.13 |
|
24 |
+
| Tsai Chin 蔡琴 | Chinese | 10.12 |
|
25 |
+
| Bruno Mars | English | 6.29 |
|
26 |
+
| Beyonce | English | 6.06 |
|
27 |
+
|
28 |
+
To make these singers sing the songs you want to listen to, just run the following commands:
|
29 |
+
|
30 |
+
### Step1: Download the checkpoint
|
31 |
+
```bash
|
32 |
+
git lfs install
|
33 |
+
git clone https://huggingface.co/amphion/singing_voice_conversion
|
34 |
+
```
|
35 |
+
|
36 |
+
### Step2: Clone the Amphion's Source Code of GitHub
|
37 |
+
```bash
|
38 |
+
git clone https://github.com/open-mmlab/Amphion.git
|
39 |
+
```
|
40 |
+
|
41 |
+
### Step3: Specify the checkpoint's path
|
42 |
+
Use the soft link to specify the downloaded checkpoint in first step:
|
43 |
+
|
44 |
+
```bash
|
45 |
+
cd Amphion
|
46 |
+
mkdir ckpts/svc
|
47 |
+
ln -s ../singing_voice_conversion/vocalist_l1_contentvec+whisper ckpts/svc/vocalist_l1_contentvec+whisper
|
48 |
+
```
|
49 |
+
|
50 |
+
### Step4: Conversion
|
51 |
+
|
52 |
+
You can follow [this recipe](https://github.com/open-mmlab/Amphion/tree/main/egs/svc/MultipleContentsSVC#4-inferenceconversion) to conduct the conversion. For example, if you want to make Taylor Swift sing the songs in the `[Your Audios Folder]``, just run:
|
53 |
+
|
54 |
+
```bash
|
55 |
+
sh egs/svc/MultipleContentsSVC/run.sh --stage 3 --gpu "0" \
|
56 |
+
--infer_expt_dir "ckpts/svc/vocalist_l1_contentvec+whisper" \
|
57 |
+
--infer_output_dir "ckpts/svc/vocalist_l1_contentvec+whisper/result" \
|
58 |
+
--infer_source_audio_dir [Your Audios Folder] \
|
59 |
+
--infer_target_speaker "vocalist_l1_TaylorSwift" \
|
60 |
+
--infer_key_shift "autoshift"
|
61 |
+
```
|
62 |
+
|
63 |
+
**Note**: The supported `infer_target_speaker` values can be seen [here](https://huggingface.co/amphion/singing_voice_conversion/blob/main/vocalist_l1_contentvec%2Bwhisper/singers.json).
|
64 |
+
|
65 |
+
## Citaions
|
66 |
+
|
67 |
+
```bibtex
|
68 |
+
@article{zhang2023leveraging,
|
69 |
+
title={Leveraging Content-based Features from Multiple Acoustic Models for Singing Voice Conversion},
|
70 |
+
author={Zhang, Xueyao and Gu, Yicheng and Chen, Haopeng and Fang, Zihao and Zou, Lexiao and Xue, Liumeng and Wu, Zhizheng},
|
71 |
+
journal={Machine Learning for Audio Worshop, NeurIPS 2023},
|
72 |
+
year={2023}
|
73 |
+
}
|
74 |
+
```
|
75 |
+
|
76 |
+
|
77 |
+
|
78 |
+
|
79 |
|
|
vocalist_l1_contentvec+whisper/data/vocalist_l1/meta_info.json
ADDED
@@ -0,0 +1,31 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
{
|
2 |
+
"dataset": "vocalist_l1",
|
3 |
+
"train": {
|
4 |
+
"size": 3180,
|
5 |
+
"hours": 6.1643
|
6 |
+
},
|
7 |
+
"test": {
|
8 |
+
"size": 114,
|
9 |
+
"hours": 0.2224
|
10 |
+
},
|
11 |
+
"singers": {
|
12 |
+
"size": 15,
|
13 |
+
"training_minutes": {
|
14 |
+
"vocalist_l1_陶喆": 45.51,
|
15 |
+
"vocalist_l1_陈奕迅": 43.36,
|
16 |
+
"vocalist_l1_汪峰": 41.08,
|
17 |
+
"vocalist_l1_李健": 38.9,
|
18 |
+
"vocalist_l1_JohnMayer": 30.83,
|
19 |
+
"vocalist_l1_Adele": 27.23,
|
20 |
+
"vocalist_l1_那英": 27.02,
|
21 |
+
"vocalist_l1_石倚洁": 24.93,
|
22 |
+
"vocalist_l1_张学友": 18.31,
|
23 |
+
"vocalist_l1_TaylorSwift": 18.31,
|
24 |
+
"vocalist_l1_王菲": 16.78,
|
25 |
+
"vocalist_l1_MichaelJackson": 15.13,
|
26 |
+
"vocalist_l1_蔡琴": 10.12,
|
27 |
+
"vocalist_l1_BrunoMars": 6.29,
|
28 |
+
"vocalist_l1_Beyonce": 6.06
|
29 |
+
}
|
30 |
+
}
|
31 |
+
}
|