Audio-to-Audio
Safetensors
torch
lucadellalib commited on
Commit
512a725
·
verified ·
1 Parent(s): 5d965c3

Upload 3 files

Browse files
Files changed (3) hide show
  1. README.md +43 -0
  2. config.json +74 -0
  3. model.safetensors +3 -0
README.md CHANGED
@@ -1,3 +1,46 @@
1
  ---
2
  license: apache-2.0
 
 
 
 
3
  ---
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
  ---
2
  license: apache-2.0
3
+ library_name: torch
4
+ base_model:
5
+ - microsoft/wavlm-large
6
+ pipeline_tag: audio-to-audio
7
  ---
8
+
9
+ # 🗜️ ZipCodec
10
+
11
+ A 6.25 Hz, 0.80 kbit/s streaming neural audio codec based on [WavLM](https://arxiv.org/abs/2110.13900).
12
+
13
+ - 📜 **Paper**: https://arxiv.org/abs/2609.11642
14
+
15
+ - 🌐 **Project Page**: https://lucadellalib.github.io/zipcodec-web/
16
+
17
+ - 💾 **GitHub**: https://github.com/lucadellalib/zipcodec
18
+
19
+ <img src="zipcodec.jpg" width="700">
20
+
21
+ ---------------------------------------------------------------------------------------------------------
22
+
23
+ ## ▶️ Quickstart
24
+
25
+ See the readme at: https://github.com/lucadellalib/zipcodec
26
+
27
+ ---------------------------------------------------------------------------------------------------------
28
+
29
+ ## @ Citing
30
+
31
+ ```bibtex
32
+ @article{dellalibera2026zipcodec,
33
+ title = {{ZipCodec}: Ultra-Low-Frame-Rate Streaming Speech Coding},
34
+ author = {Luca {Della Libera} and Cem Subakan and Mirco Ravanelli},
35
+ journal = {arXiv preprint arXiv:2609.11642},
36
+ year = {2026},
37
+ }
38
+ ```
39
+
40
+ ---------------------------------------------------------------------------------------------------------
41
+
42
+ ## 📧 Contact
43
+
44
+ [luca.dellalib@gmail.com](mailto:luca.dellalib@gmail.com)
45
+
46
+ ---------------------------------------------------------------------------------------------------------
config.json ADDED
@@ -0,0 +1,74 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "encoder_name": "LogMelSpectrogram",
3
+ "encoder_config": {
4
+ "sample_rate": 16000,
5
+ "n_fft": 400,
6
+ "win_length": 400,
7
+ "hop_length": 160,
8
+ "n_mels": 80,
9
+ "f_min": 0.0,
10
+ "f_max": 8000.0,
11
+ "eps": 1e-06,
12
+ "stft_backend": "conv"
13
+ },
14
+ "frontend_name": "Patch1d",
15
+ "frontend_config": {
16
+ "input_dim": 80,
17
+ "output_dim": 2048,
18
+ "patch_size": 16
19
+ },
20
+ "compressor_name": "ErfFormer",
21
+ "compressor_config": {
22
+ "dim": 2048,
23
+ "ffn_dim": 8192,
24
+ "num_layers": 6,
25
+ "num_heads": 16,
26
+ "num_kv_heads": 4,
27
+ "dropout": 0.0,
28
+ "window_size": 256,
29
+ "erfscale_init": 0.5,
30
+ "shift_init": 0.0,
31
+ "native_gqa": false
32
+ },
33
+ "quantizer_name": "ScalarSphericalQuantizer",
34
+ "quantizer_config": {
35
+ "input_dim": 2048,
36
+ "latent_dim": 64,
37
+ "bits_per_dim": 2,
38
+ "erfscale_init": 0.5,
39
+ "shift_init": 0.0
40
+ },
41
+ "decompressor_name": "ErfFormer",
42
+ "decompressor_config": {
43
+ "dim": 2048,
44
+ "ffn_dim": 8192,
45
+ "num_layers": 6,
46
+ "num_heads": 16,
47
+ "num_kv_heads": 4,
48
+ "dropout": 0.0,
49
+ "window_size": 256,
50
+ "erfscale_init": 0.5,
51
+ "shift_init": 0.0,
52
+ "native_gqa": false
53
+ },
54
+ "backend_name": "Unpatch1d",
55
+ "backend_config": {
56
+ "input_dim": 2048,
57
+ "output_dim": 1024,
58
+ "patch_size": 8
59
+ },
60
+ "decoder_name": "Vocos",
61
+ "decoder_config": {
62
+ "input_dim": 1024,
63
+ "num_layers": 20,
64
+ "dim": 1024,
65
+ "ffn_dim": 2048,
66
+ "kernel_size": 7,
67
+ "patch_size": 8,
68
+ "n_fft": 1024,
69
+ "win_length": 1024,
70
+ "hop_length": 320,
71
+ "istft_backend": "conv",
72
+ "eps": 1e-06
73
+ }
74
+ }
model.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:e042cb4bdb64d8bf0b9c6e8caaf3a5b20e724abf7a960a1441c0f36e9e5079c2
3
+ size 3368656896