DungeonCrawler/comfyui-audio/generate_via_api.py
Andre 458be01cde ComfyUI AudioCraft: OUTPUT_NODE Fix, API Support, HuggingFace Cache
- OUTPUT_NODE = True fuer MusicGen und AudioGen (ComfyUI API funktioniert jetzt)
- generate_via_api.py: Audio direkt ueber ComfyUI REST API generieren
- docker-compose.yml: hf_cache Volume fuer persistente Modelle nach Neustart
- Menü Musik und alle Krieger Attacken-Sounds generiert

Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
2026-03-20 18:26:57 +01:00

86 lines
3.3 KiB
Python

"""
Game Audio Generator via ComfyUI API
Aufruf: python generate_via_api.py
Voraussetzung: ComfyUI Container läuft auf localhost:8188
"""
import json
import requests
import time
import urllib.request
import os
COMFY_URL = "http://localhost:8188"
OUTPUT_DIR = os.path.join(os.path.dirname(__file__), "output", "audio")
os.makedirs(OUTPUT_DIR, exist_ok=True)
TRACKS = [
# (typ, prompt, dauer, dateiname)
("music", "dark fantasy main menu music, epic orchestral, dramatic, mysterious, dark medieval atmosphere, slow build, no vocals", 40, "menu_music"),
("sfx", "fast sword swing whoosh attack sound effect", 1, "attack_autoattack"),
("sfx", "heavy powerful sword strike crushing impact metal clang", 2, "attack_heavy_strike"),
("sfx", "massive ground slam earthquake shockwave stone cracking impact", 2, "attack_tektonischer_schlag"),
("sfx", "warrior battle cry rage grunt determination short", 2, "attack_durchbeissen"),
("sfx", "blade slash wet blood hit dark melee attack", 2, "attack_blutrausch"),
("sfx", "fast spinning whirlwind sword multiple swings whoosh", 2, "attack_wirbelwind"),
("sfx", "chains rattling metal binding restraint whoosh throw", 2, "attack_zornfesseln"),
("sfx", "berserker rage roar battle fury intense warrior scream", 3, "attack_berserker"),
]
def build_prompt(typ, prompt_text, duration, filename):
node_type = "MusicGen" if typ == "music" else "AudioGen"
widgets = [prompt_text, duration, "facebook/musicgen-small", 0] if typ == "music" else [prompt_text, duration, 0]
return {
"prompt": {
"1": {
"class_type": node_type,
"inputs": {
"prompt": prompt_text,
"duration": duration,
**({"model": "facebook/musicgen-small", "seed": 0} if typ == "music" else {"seed": 0})
}
}
},
"extra_data": {"extra_pnginfo": {"filename_hint": filename}}
}
def queue_prompt(payload):
r = requests.post(f"{COMFY_URL}/api/prompt", json=payload)
r.raise_for_status()
return r.json()["prompt_id"]
def wait_for_completion(prompt_id, timeout=300):
start = time.time()
while time.time() - start < timeout:
r = requests.get(f"{COMFY_URL}/api/history/{prompt_id}")
data = r.json()
if prompt_id in data:
return data[prompt_id]
time.sleep(2)
raise TimeoutError(f"Timeout nach {timeout}s")
def get_output_files():
r = requests.get(f"{COMFY_URL}/api/history")
return r.json()
if __name__ == "__main__":
print(f"Verbinde mit ComfyUI: {COMFY_URL}")
try:
requests.get(f"{COMFY_URL}/api/system_stats").raise_for_status()
print("ComfyUI erreichbar\n")
except Exception as e:
print(f"ComfyUI nicht erreichbar: {e}")
exit(1)
for typ, prompt, duration, filename in TRACKS:
print(f"[{typ.upper()}] {filename}: {prompt[:55]}...")
payload = build_prompt(typ, prompt, duration, filename)
try:
prompt_id = queue_prompt(payload)
print(f" queued: {prompt_id}")
result = wait_for_completion(prompt_id)
print(f" fertig!")
except Exception as e:
print(f" Fehler: {e}")
print(f"\nAlle Jobs abgeschickt. Output unter: {OUTPUT_DIR}")
print(f"Oder direkt in ComfyUI unter output/audio/")