Spaces:
Paused
Paused
Update app.py
Browse files
app.py
CHANGED
@@ -3,9 +3,9 @@ import torchaudio
|
|
3 |
from snac import SNAC
|
4 |
import gradio as gr
|
5 |
|
6 |
-
#
|
7 |
MODEL_NAME = "hubertsiuzdak/snac_24khz"
|
8 |
-
|
9 |
|
10 |
DEVICE = "cuda" if torch.cuda.is_available() else "cpu"
|
11 |
model = SNAC.from_pretrained(MODEL_NAME).eval().to(DEVICE)
|
@@ -20,18 +20,24 @@ def reconstruct(audio_in):
|
|
20 |
if data.ndim == 2:
|
21 |
data = data.mean(axis=1)
|
22 |
|
23 |
-
#
|
24 |
-
audio = torch.from_numpy(data).float().unsqueeze(0)
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
25 |
|
26 |
-
# run through SNAC
|
27 |
with torch.inference_mode():
|
28 |
audio_hat, codes = model(audio)
|
29 |
|
30 |
y = audio_hat.squeeze().cpu().numpy()
|
31 |
-
return (
|
32 |
|
33 |
with gr.Blocks(title="SNAC Round-Trip Demo") as demo:
|
34 |
-
gr.Markdown("## 🎧 SNAC Audio Reconstructor (
|
35 |
|
36 |
with gr.Row():
|
37 |
with gr.Column():
|
|
|
3 |
from snac import SNAC
|
4 |
import gradio as gr
|
5 |
|
6 |
+
# choose your SNAC model + target sample rate
|
7 |
MODEL_NAME = "hubertsiuzdak/snac_24khz"
|
8 |
+
TARGET_SR = 24000
|
9 |
|
10 |
DEVICE = "cuda" if torch.cuda.is_available() else "cpu"
|
11 |
model = SNAC.from_pretrained(MODEL_NAME).eval().to(DEVICE)
|
|
|
20 |
if data.ndim == 2:
|
21 |
data = data.mean(axis=1)
|
22 |
|
23 |
+
# torchify
|
24 |
+
audio = torch.from_numpy(data).float().unsqueeze(0) # [1,T]
|
25 |
+
|
26 |
+
# resample to target SR
|
27 |
+
if sr != TARGET_SR:
|
28 |
+
audio = torchaudio.functional.resample(audio, orig_freq=sr, new_freq=TARGET_SR)
|
29 |
+
|
30 |
+
# expand to [B,1,T]
|
31 |
+
audio = audio.unsqueeze(0).to(DEVICE)
|
32 |
|
|
|
33 |
with torch.inference_mode():
|
34 |
audio_hat, codes = model(audio)
|
35 |
|
36 |
y = audio_hat.squeeze().cpu().numpy()
|
37 |
+
return (TARGET_SR, y)
|
38 |
|
39 |
with gr.Blocks(title="SNAC Round-Trip Demo") as demo:
|
40 |
+
gr.Markdown("## 🎧 SNAC Audio Reconstructor (with resampling)")
|
41 |
|
42 |
with gr.Row():
|
43 |
with gr.Column():
|