HeartMuLa commited on
Commit
430ca55
·
verified ·
1 Parent(s): 9d0a8a5

Prepare encoder release documentation

Browse files
Files changed (1) hide show
  1. README.md +10 -11
README.md CHANGED
@@ -13,32 +13,31 @@ with [HeartCodec-oss-20260123](https://huggingface.co/HeartMuLa/HeartCodec-oss-2
13
  which supplies the shared RVQ quantizer and audio decoder. Together they support
14
  audio → eight token streams at 12.5 Hz → 48 kHz stereo reconstruction.
15
 
16
- This repository is a private development preview. The matching code is in
17
- [HeartMuLa/heartlib-codec-dev](https://github.com/HeartMuLa/heartlib-codec-dev);
18
- access to both repositories is required. The encoder loading API has not yet
19
- been merged into the public `HeartMuLa/heartlib` repository.
 
20
 
21
  ## Installation and reconstruction
22
 
23
  Use Python 3.10, a CUDA-compatible PyTorch installation, and `ffmpeg` for MP3
24
- output. Install this development copy of `heartlib` in a separate environment;
25
- do not install both the public and development copies into one environment.
26
 
27
  ```bash
28
- git clone https://github.com/HeartMuLa/heartlib-codec-dev.git
29
- cd heartlib-codec-dev
30
  pip install -e .
31
- hf auth login
32
  hf download HeartMuLa/HeartCodec-oss-encoder --local-dir ./ckpt/HeartCodec-oss-encoder
33
  hf download HeartMuLa/HeartCodec-oss-20260123 --local-dir ./ckpt/HeartCodec-oss-20260123
34
  python examples/run_music_reconstruction.py \
35
  --encoder_path ./ckpt/HeartCodec-oss-encoder \
36
  --decoder_path ./ckpt/HeartCodec-oss-20260123 \
37
- --input_path /path/to/your/audio.mp3 \
38
  --save_path ./reconstructed.mp3
39
  ```
40
 
41
- Use authorized input audio. MP3 is encoded directly from float32 output at
42
  320 kbps, without an intermediate PCM WAV. WAV output is available by selecting
43
  a `.wav` output path. Output is trimmed to the input duration.
44
 
 
13
  which supplies the shared RVQ quantizer and audio decoder. Together they support
14
  audio → eight token streams at 12.5 Hz → 48 kHz stereo reconstruction.
15
 
16
+ The matching code and reconstruction examples are in
17
+ [HeartMuLa/heartlib](https://github.com/HeartMuLa/heartlib).
18
+
19
+ You can now encode your own audio into tokens for fine-tuning HeartMuLa.
20
+ We hope this contribution will be useful to the music research community.
21
 
22
  ## Installation and reconstruction
23
 
24
  Use Python 3.10, a CUDA-compatible PyTorch installation, and `ffmpeg` for MP3
25
+ output. Install `heartlib` in a separate environment.
 
26
 
27
  ```bash
28
+ git clone https://github.com/HeartMuLa/heartlib.git
29
+ cd heartlib
30
  pip install -e .
 
31
  hf download HeartMuLa/HeartCodec-oss-encoder --local-dir ./ckpt/HeartCodec-oss-encoder
32
  hf download HeartMuLa/HeartCodec-oss-20260123 --local-dir ./ckpt/HeartCodec-oss-20260123
33
  python examples/run_music_reconstruction.py \
34
  --encoder_path ./ckpt/HeartCodec-oss-encoder \
35
  --decoder_path ./ckpt/HeartCodec-oss-20260123 \
36
+ --input_path ./assets/reference.mp3 \
37
  --save_path ./reconstructed.mp3
38
  ```
39
 
40
+ Replace `--input_path` to use your own authorized audio. MP3 is encoded directly from float32 output at
41
  320 kbps, without an intermediate PCM WAV. WAV output is available by selecting
42
  a `.wav` output path. Output is trimmed to the input duration.
43