1580158
/
ComfyUI
mirror of https://github.com/comfyanonymous/ComfyUI
1
0
Fork 0
Code Issues Packages Projects Releases Wiki Activity

Fix LTXV Reference Audio node (#13531)

pull/11377/merge
Jukka Seppänen 2026-04-23 21:13:17 +03:00 committed by GitHub
parent abf3d56f27
commit 6fbb6b6f49
No known key found for this signature in database
GPG Key ID: B5690EEEBB952194
1 changed files with 9 additions and 1 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

10
comfy_extras/nodes_lt.py
View File

@ -1,6 +1,7 @@
import nodes import nodes
import node_helpers import node_helpers
import torch import torch
import torchaudio
import comfy.model_management import comfy.model_management
import comfy.model_sampling import comfy.model_sampling
import comfy.samplers import comfy.samplers
@ -711,7 +712,14 @@ class LTXVReferenceAudio(io.ComfyNode):
@classmethod @classmethod
def execute(cls, model, positive, negative, reference_audio, audio_vae, identity_guidance_scale, start_percent, end_percent) -> io.NodeOutput: def execute(cls, model, positive, negative, reference_audio, audio_vae, identity_guidance_scale, start_percent, end_percent) -> io.NodeOutput:
# Encode reference audio to latents and patchify # Encode reference audio to latents and patchify
audio_latents = audio_vae.encode(reference_audio) sample_rate = reference_audio["sample_rate"]
vae_sample_rate = getattr(audio_vae, "audio_sample_rate", 44100)
if vae_sample_rate != sample_rate:
waveform = torchaudio.functional.resample(reference_audio["waveform"], sample_rate, vae_sample_rate)
else:
waveform = reference_audio["waveform"]
audio_latents = audio_vae.encode(waveform.movedim(1, -1))
b, c, t, f = audio_latents.shape b, c, t, f = audio_latents.shape
ref_tokens = audio_latents.permute(0, 2, 1, 3).reshape(b, t, c * f) ref_tokens = audio_latents.permute(0, 2, 1, 3).reshape(b, t, c * f)
ref_audio = {"tokens": ref_tokens} ref_audio = {"tokens": ref_tokens}