Skip to content

The most powerful and modular diffusion model GUI, api and backend with a graph/nodes interface. The fastest local inference engine in the world.

Resample audio to 44100 when VAE encoding it.

comfyanonymouseab211bb1e79
diff --git a/comfy_extras/nodes_audio.py b/comfy_extras/nodes_audio.py
index 5f4bd3540..57d0a20ab 100644
--- a/comfy_extras/nodes_audio.py
+++ b/comfy_extras/nodes_audio.py
@@ -31,7 +31,13 @@ class VAEEncodeAudio:
CATEGORY = "_for_testing/audio"
def encode(self, vae, audio):
- t = vae.encode(audio["waveform"].movedim(1, -1))
+ sample_rate = audio["sample_rate"]
+ if 44100 != sample_rate:
+ waveform = torchaudio.functional.resample(audio["waveform"], sample_rate, 44100)
+ else:
+ waveform = audio["waveform"]
+
+ t = vae.encode(waveform.movedim(1, -1))
return ({"samples":t}, )
class VAEDecodeAudio: