Andre
f3c895ac43
- comfyui_audiocraft: eigene MusicGen + AudioGen Nodes - workflow_game_audio.json: vorgefertigter Workflow fuer Dungeon Musik und Sound Effects Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
92 lines
2.9 KiB
Python
92 lines
2.9 KiB
Python
import torch
|
|
import torchaudio
|
|
import os
|
|
import time
|
|
|
|
|
|
class MusicGenNode:
|
|
@classmethod
|
|
def INPUT_TYPES(cls):
|
|
return {
|
|
"required": {
|
|
"prompt": ("STRING", {"multiline": True, "default": "dark dungeon ambience, slow, mysterious"}),
|
|
"duration": ("FLOAT", {"default": 10.0, "min": 1.0, "max": 60.0, "step": 1.0}),
|
|
"model": (["facebook/musicgen-small", "facebook/musicgen-medium", "facebook/musicgen-large", "facebook/musicgen-stereo-medium"],),
|
|
"seed": ("INT", {"default": 0, "min": 0, "max": 2**32 - 1}),
|
|
}
|
|
}
|
|
|
|
RETURN_TYPES = ("AUDIO_PATH",)
|
|
RETURN_NAMES = ("audio_file",)
|
|
FUNCTION = "generate"
|
|
CATEGORY = "AudioCraft"
|
|
|
|
def generate(self, prompt, duration, model, seed):
|
|
from audiocraft.models import MusicGen
|
|
from audiocraft.data.audio import audio_write
|
|
|
|
torch.manual_seed(seed)
|
|
|
|
print(f"[MusicGen] Loading model: {model}")
|
|
mg = MusicGen.get_pretrained(model)
|
|
mg.set_generation_params(duration=duration)
|
|
|
|
print(f"[MusicGen] Generating: {prompt}")
|
|
wav = mg.generate([prompt])
|
|
|
|
output_dir = "/app/ComfyUI/output/audio"
|
|
os.makedirs(output_dir, exist_ok=True)
|
|
filename = f"musicgen_{int(time.time())}"
|
|
out_path = os.path.join(output_dir, filename)
|
|
audio_write(out_path, wav[0].cpu(), mg.sample_rate, strategy="loudness")
|
|
|
|
return (out_path + ".wav",)
|
|
|
|
|
|
class AudioGenNode:
|
|
@classmethod
|
|
def INPUT_TYPES(cls):
|
|
return {
|
|
"required": {
|
|
"prompt": ("STRING", {"multiline": True, "default": "sword clash metal sound effect"}),
|
|
"duration": ("FLOAT", {"default": 3.0, "min": 0.5, "max": 30.0, "step": 0.5}),
|
|
"seed": ("INT", {"default": 0, "min": 0, "max": 2**32 - 1}),
|
|
}
|
|
}
|
|
|
|
RETURN_TYPES = ("AUDIO_PATH",)
|
|
RETURN_NAMES = ("audio_file",)
|
|
FUNCTION = "generate"
|
|
CATEGORY = "AudioCraft"
|
|
|
|
def generate(self, prompt, duration, seed):
|
|
from audiocraft.models import AudioGen
|
|
from audiocraft.data.audio import audio_write
|
|
|
|
torch.manual_seed(seed)
|
|
|
|
print(f"[AudioGen] Loading model...")
|
|
ag = AudioGen.get_pretrained("facebook/audiogen-medium")
|
|
ag.set_generation_params(duration=duration)
|
|
|
|
print(f"[AudioGen] Generating: {prompt}")
|
|
wav = ag.generate([prompt])
|
|
|
|
output_dir = "/app/ComfyUI/output/audio"
|
|
os.makedirs(output_dir, exist_ok=True)
|
|
filename = f"audiogen_{int(time.time())}"
|
|
out_path = os.path.join(output_dir, filename)
|
|
audio_write(out_path, wav[0].cpu(), ag.sample_rate, strategy="loudness")
|
|
|
|
return (out_path + ".wav",)
|
|
|
|
|
|
NODE_CLASS_MAPPINGS = {
|
|
"MusicGen": MusicGenNode,
|
|
"AudioGen": AudioGenNode,
|
|
}
|
|
|
|
NODE_DISPLAY_NAME_MAPPINGS = {
|
|
"MusicGen": "MusicGen (Musik)",
|
|
"AudioGen": "AudioGen (Sound Effects)",
|
|
}
|