Update README.md
Browse files
README.md
CHANGED
|
@@ -5,7 +5,7 @@ base_model:
|
|
| 5 |
- speechbrain/tts-hifigan-unit-hubert-l6-k100-ljspeech
|
| 6 |
library_name: speechbrain
|
| 7 |
---
|
| 8 |
-
# Basque Unit-HiFiGAN Vocoder (Voices:
|
| 9 |
## Model Summary
|
| 10 |
|
| 11 |
This repository provides a Unit-HiFiGAN vocoder trained to synthesize high-fidelity Basque speech from discrete HuBERT-derived unit sequences. The model supports two speaker identities, Maider and Antton, using learned speaker-conditioning embeddings. It is compatible with HuBERT features extracted from layer 9 and clustered using a KMeans (k=1000) quantizer.
|
|
@@ -49,11 +49,11 @@ DEVICE = "cuda" if torch.cuda.is_available() else "cpu"
|
|
| 49 |
SR = 16000
|
| 50 |
|
| 51 |
# 1. Load HuBERT
|
| 52 |
-
processor = Wav2Vec2Processor.from_pretrained("
|
| 53 |
-
hubert = HubertModel.from_pretrained("
|
| 54 |
|
| 55 |
# 2. Load KMeans
|
| 56 |
-
kmeans_path = hf_hub_download("
|
| 57 |
kmeans = joblib.load(kmeans_path)
|
| 58 |
|
| 59 |
# 3. Load vocoder
|
|
|
|
| 5 |
- speechbrain/tts-hifigan-unit-hubert-l6-k100-ljspeech
|
| 6 |
library_name: speechbrain
|
| 7 |
---
|
| 8 |
+
# Basque Unit-HiFiGAN Vocoder (Voices: Maider & Antton)
|
| 9 |
## Model Summary
|
| 10 |
|
| 11 |
This repository provides a Unit-HiFiGAN vocoder trained to synthesize high-fidelity Basque speech from discrete HuBERT-derived unit sequences. The model supports two speaker identities, Maider and Antton, using learned speaker-conditioning embeddings. It is compatible with HuBERT features extracted from layer 9 and clustered using a KMeans (k=1000) quantizer.
|
|
|
|
| 49 |
SR = 16000
|
| 50 |
|
| 51 |
# 1. Load HuBERT
|
| 52 |
+
processor = Wav2Vec2Processor.from_pretrained("Ansu/HiFiGAN-Basque-Maider-Antton")
|
| 53 |
+
hubert = HubertModel.from_pretrained("Ansu/HiFiGAN-Basque-Maider-Antton").to(DEVICE).eval()
|
| 54 |
|
| 55 |
# 2. Load KMeans
|
| 56 |
+
kmeans_path = hf_hub_download("Ansu/HiFiGAN-Basque-Maider-Antton", "kmeans/basque_hubert_k1000_L9.pt")
|
| 57 |
kmeans = joblib.load(kmeans_path)
|
| 58 |
|
| 59 |
# 3. Load vocoder
|