#!/usr/bin/env python3
"""
Start RTMPS stream to X/Twitter using MuseTalk neural lip sync
with the cyborg Farnsworth avatar.
MuseTalk provides photorealistic audio-driven face animation at 30+ FPS
on V100 GPU, replacing the mesh-warping FaceRigAnimator.
Prerequisites:
bash scripts/setup_musetalk.sh # Install MuseTalk + download weights
export TWITTER_STREAM_KEY=... # Set your X stream key
"""
import asyncio
import os
import sys
from pathlib import Path
sys.path.insert(0, str(Path(__file__).parent.parent))
from loguru import logger
logger.remove()
logger.add(
sys.stderr,
format="<green>{time:HH:mm:ss}</green> | <level>{level: <8}</level> | "
"<cyan>{name}</cyan> - <level>{message}</level>",
level="INFO",
)
logger.add("/tmp/vtuber_musetalk.log", rotation="50 MB", level="DEBUG")
async def main():
from farnsworth.integration.vtuber.vtuber_core import FarnsworthVTuber, VTuberConfig
from farnsworth.integration.vtuber.stream_manager import StreamQuality
from farnsworth.integration.vtuber.avatar_controller import AvatarBackend
# Load stream key
stream_key = os.environ.get("TWITTER_STREAM_KEY", "")
if not stream_key:
logger.error("TWITTER_STREAM_KEY not set in environment")
sys.exit(1)
face_image = "/workspace/Farnsworth/assets/farnsworth_closeup.png"
if not Path(face_image).exists():
logger.error(f"Face image not found: {face_image}")
sys.exit(1)
proxy_face = "/workspace/MuseTalk/assets/demo/man/man.png"
musetalk_dir = "/workspace/MuseTalk"
if not Path(musetalk_dir).exists():
logger.error(
f"MuseTalk not found at {musetalk_dir}. "
f"Run: bash scripts/setup_musetalk.sh"
)
sys.exit(1)
logger.info("=" * 50)
logger.info(" FARNSWORTH RTMPS STREAM TO X")
logger.info(" Backend: MUSETALK (neural lip sync)")
logger.info(f" Avatar: {face_image}")
logger.info(f" MuseTalk: {musetalk_dir}")
logger.info(f" Proxy face: {proxy_face}")
logger.info(f" Stream key: {stream_key[:4]}...")
logger.info("=" * 50)
config = VTuberConfig(
name="Farnsworth",
persona="An eccentric AI scientist leading a collective of AI agents",
avatar_backend=AvatarBackend.MUSETALK,
avatar_face_image=face_image,
musetalk_dir=musetalk_dir,
musetalk_version="v15",
musetalk_proxy_face=proxy_face,
avatar_width=854,
avatar_height=480,
avatar_fps=25, # MuseTalk native fps
stream_key=stream_key,
stream_quality=StreamQuality.MEDIUM,
stream_platform="twitter",
simulate_chat=False,
use_swarm_collective=True,
swarm_agents=["Farnsworth", "Grok", "DeepSeek", "Gemini"],
enable_chat=True,
idle_chat_interval=30.0,
)
vtuber = FarnsworthVTuber(config)
try:
logger.info("Starting VTuber stream with MuseTalk...")
success = await vtuber.start()
if not success:
logger.error("Failed to start stream")
sys.exit(1)
logger.info("=" * 50)
logger.info(" VTUBER IS LIVE ON X (MuseTalk)")
logger.info("=" * 50)
logger.info("Press Ctrl+C to stop")
while vtuber.is_live:
await asyncio.sleep(5)
stats = vtuber.stats
if stats.get("stream_stats"):
fps = stats["stream_stats"].get("avg_fps", 0)
frames = stats["stream_stats"].get("frames_sent", 0)
logger.info(f"Stream: {fps:.1f} fps | {frames} frames sent")
except KeyboardInterrupt:
logger.info("\nShutdown requested...")
finally:
await vtuber.stop()
logger.info("Stream stopped.")
if __name__ == "__main__":
asyncio.run(main())
