feat: initial release - Audio ↔ Tensor bridge nodes for ComfyUI

- AudioToTensor: Converts ComfyUI AUDIO to VCAUDIOTENSOR (mono, resampled) - TensorToAudio: Converts VCAUDIOTENSOR back to ComfyUI AUDIO - Compatible with MuseTalk-KJ, VoiceCraft, and any Whisper-based node - Zero additional dependencies (uses PyTorch + torchaudio)
2026-02-20 15:41:32 -03:00
commit 9ebd2a85e5
6 changed files with 205 additions and 0 deletions
@@ -0,0 +1,27 @@
+# Python
+__pycache__/
+*.py[cod]
+*$py.class
+*.so
+*.egg-info/
+dist/
+build/
+*.egg
+
+# Virtual environments
+.venv/
+venv/
+env/
+
+# IDE
+.vscode/
+.idea/
+*.swp
+*.swo
+*~
+.DS_Store
+
+# Testing
+.pytest_cache/
+.coverage
+htmlcov/
@@ -0,0 +1,21 @@
+MIT License
+
+Copyright (c) 2026 Ryan Holanda
+
+Permission is hereby granted, free of charge, to any person obtaining a copy
+of this software and associated documentation files (the "Software"), to deal
+in the Software without restriction, including without limitation the rights
+to use, copy, modify, merge, publish, distribute, sublicense, and/or sell
+copies of the Software, and to permit persons to whom the Software is
+furnished to do so, subject to the following conditions:
+
+The above copyright notice and this permission notice shall be included in all
+copies or substantial portions of the Software.
+
+THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
+AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
+LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
+OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
+SOFTWARE.
@@ -0,0 +1,29 @@
+# ComfyUI-AudioBridge
+
+Bridges ComfyUI's native **AUDIO** type to raw audio tensors (**VCAUDIOTENSOR**) used by nodes like [MuseTalk-KJ](https://github.com/kijai/ComfyUI-MuseTalk-KJ) and VoiceCraft.
+
+## Nodes
+
+| Node               | Description                                                               |
+| ------------------ | ------------------------------------------------------------------------- |
+| **Audio → Tensor** | Converts `AUDIO` → `VCAUDIOTENSOR` (mono, resampled to target rate)       |
+| **Tensor → Audio** | Converts `VCAUDIOTENSOR` → `AUDIO` (wraps tensor back for standard nodes) |
+
+## Why?
+
+ComfyUI's native `AUDIO` type is a dict (`{waveform, sample_rate}`), but some custom nodes expect `VHS_AUDIO` (a callable) or `VCAUDIOTENSOR` (a raw torch tensor). This node bridges the gap without requiring VideoHelperSuite or duplicate audio loading.
+
+## Install
+
+```bash
+cd ComfyUI/custom_nodes
+git clone https://github.com/RyanHolanda/ComfyUI-AudioBridge.git
+```
+
+No additional dependencies required — uses PyTorch and torchaudio (already included with ComfyUI).
+
+## Usage Example
+
+```
+LoadAudio → [Audio → Tensor] → whisper_to_features → MuseTalk Sampler
+```
@@ -0,0 +1,2 @@
+from .nodes import NODE_CLASS_MAPPINGS, NODE_DISPLAY_NAME_MAPPINGS
+__all__ = ["NODE_CLASS_MAPPINGS", "NODE_DISPLAY_NAME_MAPPINGS"]
@@ -0,0 +1,104 @@
+"""
+ComfyUI-AudioBridge
+Bridges ComfyUI's native AUDIO type to other audio formats used by
+custom nodes (e.g. MuseTalk, VoiceCraft).
+"""
+import torch
+import torchaudio
+
+
+class AudioToTensor:
+    """
+    Converts ComfyUI's native AUDIO type to a raw audio tensor.
+    
+    ComfyUI AUDIO = {"waveform": Tensor, "sample_rate": int}
+    Output = mono torch tensor resampled to target rate (default 16kHz).
+    
+    Compatible with MuseTalk-KJ's VCAUDIOTENSOR input type and any node
+    that expects a raw audio waveform tensor.
+    """
+
+    @classmethod
+    def INPUT_TYPES(s):
+        return {
+            "required": {
+                "audio": ("AUDIO",),
+                "target_sample_rate": ("INT", {
+                    "default": 16000,
+                    "min": 8000,
+                    "max": 48000,
+                    "step": 1000,
+                    "tooltip": "Target sample rate in Hz. 16000 for Whisper/MuseTalk."
+                }),
+            },
+        }
+
+    RETURN_TYPES = ("VCAUDIOTENSOR", "INT")
+    RETURN_NAMES = ("audio_tensor", "audio_dur")
+    FUNCTION = "convert"
+    CATEGORY = "AudioBridge"
+    DESCRIPTION = "Converts ComfyUI AUDIO to a raw tensor (VCAUDIOTENSOR). Use with MuseTalk, VoiceCraft, or any node expecting raw audio tensors."
+
+    def convert(self, audio, target_sample_rate):
+        waveform = audio["waveform"].squeeze(0)  # (channels, samples)
+        sample_rate = audio["sample_rate"]
+
+        # Convert to mono
+        if waveform.shape[0] > 1:
+            waveform = waveform.mean(0, keepdim=True)
+
+        # Resample if needed
+        if sample_rate != target_sample_rate:
+            waveform = torchaudio.functional.resample(
+                waveform, sample_rate, target_sample_rate
+            )
+
+        audio_dur = waveform.shape[1] / target_sample_rate
+        return (waveform, int(audio_dur))
+
+
+class TensorToAudio:
+    """
+    Converts a raw audio tensor back to ComfyUI's native AUDIO type.
+    Useful for routing processed audio back into standard ComfyUI nodes.
+    """
+
+    @classmethod
+    def INPUT_TYPES(s):
+        return {
+            "required": {
+                "audio_tensor": ("VCAUDIOTENSOR",),
+                "sample_rate": ("INT", {
+                    "default": 16000,
+                    "min": 8000,
+                    "max": 48000,
+                    "step": 1000,
+                }),
+            },
+        }
+
+    RETURN_TYPES = ("AUDIO",)
+    RETURN_NAMES = ("audio",)
+    FUNCTION = "convert"
+    CATEGORY = "AudioBridge"
+    DESCRIPTION = "Converts a raw audio tensor (VCAUDIOTENSOR) back to ComfyUI AUDIO type."
+
+    def convert(self, audio_tensor, sample_rate):
+        # Ensure correct shape: (batch, channels, samples)
+        if audio_tensor.dim() == 1:
+            audio_tensor = audio_tensor.unsqueeze(0).unsqueeze(0)
+        elif audio_tensor.dim() == 2:
+            audio_tensor = audio_tensor.unsqueeze(0)
+
+        return ({"waveform": audio_tensor, "sample_rate": sample_rate},)
+
+
+NODE_CLASS_MAPPINGS = {
+    "AudioToTensor": AudioToTensor,
+    "TensorToAudio": TensorToAudio,
+}
+
+NODE_DISPLAY_NAME_MAPPINGS = {
+    "AudioToTensor": "Audio → Tensor",
+    "TensorToAudio": "Tensor → Audio",
+}
@@ -0,0 +1,22 @@
+[project]
+name = "comfyui-audiobridge"
+version = "1.0.0"
+description = "Bridges ComfyUI's native AUDIO type to raw audio tensors (VCAUDIOTENSOR) for MuseTalk, VoiceCraft, and more."
+readme = "README.md"
+license = {text = "MIT"}
+requires-python = ">=3.10"
+authors = [
+    {name = "Ryan Holanda", email = ""},
+]
+keywords = ["comfyui", "audio", "musetalk", "whisper", "tts"]
+classifiers = [
+    "Development Status :: 5 - Production/Stable",
+    "License :: OSI Approved :: MIT License",
+    "Programming Language :: Python :: 3",
+    "Topic :: Multimedia :: Sound/Audio",
+]
+
+[project.urls]
+Homepage = "https://github.com/RyanHolanda/ComfyUI-AudioBridge"
+Repository = "https://github.com/RyanHolanda/ComfyUI-AudioBridge"
+Issues = "https://github.com/RyanHolanda/ComfyUI-AudioBridge/issues"