Initial commit: AbletonMCP-AI complete system
- MCP Server with audio fallback, sample management - Song generator with bus routing - Reference listener and audio resampler - Vector-based sample search - Master chain with limiter and calibration - Fix: Audio fallback now works without M4L - Fix: Full song detection in sample loader Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
This commit is contained in:
renato97
264
AbletonMCP_AI/MCP_Server/reference_stem_builder.py
Normal file
264
AbletonMCP_AI/MCP_Server/reference_stem_builder.py
Normal file
@@ -0,0 +1,264 @@
|
||||
"""
|
||||
reference_stem_builder.py - Rebuild an Ableton arrangement directly from a reference track.
|
||||
"""
|
||||
|
||||
from __future__ import annotations
|
||||
|
||||
import json
|
||||
import logging
|
||||
import socket
|
||||
from pathlib import Path
|
||||
from typing import Any, Dict, List, Tuple
|
||||
|
||||
import soundfile as sf
|
||||
import torch
|
||||
from demucs.apply import apply_model
|
||||
from demucs.pretrained import get_model
|
||||
|
||||
try:
|
||||
import librosa
|
||||
except ImportError: # pragma: no cover
|
||||
librosa = None
|
||||
|
||||
try:
|
||||
from reference_listener import ReferenceAudioListener
|
||||
except ImportError: # pragma: no cover
|
||||
from .reference_listener import ReferenceAudioListener
|
||||
|
||||
|
||||
logger = logging.getLogger("ReferenceStemBuilder")
|
||||
|
||||
HOST = "127.0.0.1"
|
||||
PORT = 9877
|
||||
MESSAGE_TERMINATOR = b"\n"
|
||||
SCRIPT_DIR = Path(__file__).resolve().parent
|
||||
PACKAGE_DIR = SCRIPT_DIR.parent
|
||||
PROJECT_SAMPLES_DIR = PACKAGE_DIR.parent / "librerias" / "organized_samples"
|
||||
SAMPLES_DIR = str(PROJECT_SAMPLES_DIR)
|
||||
|
||||
TRACK_LAYOUT = (
|
||||
("REFERENCE FULL", 59, 0.72, True),
|
||||
("REF DRUMS", 10, 0.84, False),
|
||||
("REF BASS", 30, 0.82, False),
|
||||
("REF OTHER", 50, 0.68, False),
|
||||
("REF VOCALS", 40, 0.70, False),
|
||||
)
|
||||
|
||||
SECTION_BLUEPRINTS = {
|
||||
"club": [
|
||||
("INTRO DJ", 16),
|
||||
("GROOVE A", 16),
|
||||
("VOCAL BUILD", 8),
|
||||
("DROP A", 16),
|
||||
("BREAKDOWN", 8),
|
||||
("BUILD B", 8),
|
||||
("DROP B", 16),
|
||||
("PEAK", 8),
|
||||
("OUTRO DJ", 16),
|
||||
],
|
||||
"standard": [
|
||||
("INTRO", 8),
|
||||
("BUILD", 8),
|
||||
("DROP A", 16),
|
||||
("BREAK", 8),
|
||||
("DROP B", 16),
|
||||
("OUTRO", 8),
|
||||
],
|
||||
}
|
||||
|
||||
|
||||
class AbletonSocketClient:
|
||||
def __init__(self, host: str = HOST, port: int = PORT):
|
||||
self.host = host
|
||||
self.port = port
|
||||
|
||||
def send(self, command_type: str, params: Dict[str, Any] | None = None, timeout: float = 30.0) -> Dict[str, Any]:
|
||||
payload = json.dumps({"type": command_type, "params": params or {}}, separators=(",", ":")).encode("utf-8") + MESSAGE_TERMINATOR
|
||||
with socket.create_connection((self.host, self.port), timeout=timeout) as sock:
|
||||
sock.sendall(payload)
|
||||
data = b""
|
||||
while not data.endswith(MESSAGE_TERMINATOR):
|
||||
chunk = sock.recv(65536)
|
||||
if not chunk:
|
||||
break
|
||||
data += chunk
|
||||
if not data:
|
||||
raise RuntimeError(f"Sin respuesta para {command_type}")
|
||||
return json.loads(data.decode("utf-8", errors="replace").strip())
|
||||
|
||||
|
||||
def _resolve_reference_profile(reference_path: Path) -> Dict[str, Any]:
|
||||
listener = ReferenceAudioListener(SAMPLES_DIR)
|
||||
analysis = listener.analyze_reference(str(reference_path))
|
||||
structure = "club" if analysis.get("duration", 0.0) >= 180 else "standard"
|
||||
return {
|
||||
"tempo": float(analysis.get("tempo", 128.0) or 128.0),
|
||||
"key": str(analysis.get("key", "") or ""),
|
||||
"duration": float(analysis.get("duration", 0.0) or 0.0),
|
||||
"structure": structure,
|
||||
"listener_device": analysis.get("device", "cpu"),
|
||||
}
|
||||
|
||||
|
||||
def ensure_reference_wav(reference_path: Path) -> Path:
|
||||
if reference_path.suffix.lower() == ".wav":
|
||||
return reference_path
|
||||
|
||||
if librosa is None:
|
||||
raise RuntimeError("librosa no está disponible para convertir la referencia a WAV")
|
||||
|
||||
wav_path = reference_path.with_suffix(".wav")
|
||||
if wav_path.exists() and wav_path.stat().st_size > 0:
|
||||
return wav_path
|
||||
|
||||
y, sr = librosa.load(str(reference_path), sr=44100, mono=False)
|
||||
if y.ndim == 1:
|
||||
y = y.reshape(1, -1)
|
||||
sf.write(str(wav_path), y.T, sr, subtype="PCM_16")
|
||||
return wav_path
|
||||
|
||||
|
||||
def separate_stems(reference_wav: Path, output_dir: Path) -> Dict[str, Path]:
|
||||
output_dir.mkdir(parents=True, exist_ok=True)
|
||||
stem_root = output_dir / reference_wav.stem
|
||||
expected = {
|
||||
"reference": reference_wav,
|
||||
"drums": stem_root / "drums.wav",
|
||||
"bass": stem_root / "bass.wav",
|
||||
"other": stem_root / "other.wav",
|
||||
"vocals": stem_root / "vocals.wav",
|
||||
}
|
||||
if all(path.exists() and path.stat().st_size > 0 for path in expected.values()):
|
||||
return expected
|
||||
|
||||
audio, sr = sf.read(str(reference_wav), always_2d=True)
|
||||
if sr != 44100:
|
||||
raise RuntimeError(f"Sample rate inesperado en referencia WAV: {sr}")
|
||||
|
||||
model = get_model("htdemucs")
|
||||
model.cpu()
|
||||
model.eval()
|
||||
waveform = torch.tensor(audio.T, dtype=torch.float32)
|
||||
separated = apply_model(model, waveform[None], device="cpu", progress=False)[0]
|
||||
|
||||
stem_root.mkdir(parents=True, exist_ok=True)
|
||||
for stem_name, tensor in zip(model.sources, separated):
|
||||
stem_path = stem_root / f"{stem_name}.wav"
|
||||
sf.write(str(stem_path), tensor.detach().cpu().numpy().T, sr, subtype="PCM_16")
|
||||
|
||||
return expected
|
||||
|
||||
|
||||
def _sections_for_structure(structure: str) -> List[Tuple[str, int]]:
|
||||
return list(SECTION_BLUEPRINTS.get(structure.lower(), SECTION_BLUEPRINTS["standard"]))
|
||||
|
||||
|
||||
def _create_track(client: AbletonSocketClient, name: str, color: int, volume: float) -> int:
|
||||
response = client.send("create_track", {"type": "audio", "index": -1})
|
||||
if response.get("status") != "success":
|
||||
raise RuntimeError(response.get("message", f"No se pudo crear {name}"))
|
||||
track_index = int(response.get("result", {}).get("index"))
|
||||
client.send("set_track_name", {"index": track_index, "name": name})
|
||||
client.send("set_track_color", {"index": track_index, "color": color})
|
||||
client.send("set_track_volume", {"index": track_index, "volume": volume})
|
||||
return track_index
|
||||
|
||||
|
||||
def _import_full_length_audio(client: AbletonSocketClient, track_index: int, file_path: Path, name: str) -> None:
|
||||
response = client.send("create_arrangement_audio_pattern", {
|
||||
"track_index": track_index,
|
||||
"file_path": str(file_path),
|
||||
"positions": [0.0],
|
||||
"name": name,
|
||||
}, timeout=120.0)
|
||||
if response.get("status") != "success":
|
||||
raise RuntimeError(response.get("message", f"No se pudo importar {name}"))
|
||||
|
||||
|
||||
def _prepare_navigation_scenes(client: AbletonSocketClient, structure: str) -> None:
|
||||
sections = _sections_for_structure(structure)
|
||||
session_info = client.send("get_session_info")
|
||||
if session_info.get("status") != "success":
|
||||
return
|
||||
|
||||
scene_count = int(session_info.get("result", {}).get("num_scenes", 0) or 0)
|
||||
target_count = len(sections)
|
||||
|
||||
while scene_count < target_count:
|
||||
create_response = client.send("create_scene", {"index": -1})
|
||||
if create_response.get("status") != "success":
|
||||
break
|
||||
scene_count += 1
|
||||
|
||||
while scene_count > target_count and scene_count > 1:
|
||||
delete_response = client.send("delete_scene", {"index": scene_count - 1})
|
||||
if delete_response.get("status") != "success":
|
||||
break
|
||||
scene_count -= 1
|
||||
|
||||
for scene_index, (section_name, _) in enumerate(sections):
|
||||
client.send("set_scene_name", {"index": scene_index, "name": section_name})
|
||||
|
||||
|
||||
def rebuild_project_from_reference(reference_path: Path) -> Dict[str, Any]:
|
||||
reference_path = reference_path.resolve()
|
||||
if not reference_path.exists():
|
||||
raise FileNotFoundError(reference_path)
|
||||
|
||||
profile = _resolve_reference_profile(reference_path)
|
||||
reference_wav = ensure_reference_wav(reference_path)
|
||||
stems = separate_stems(reference_wav, reference_path.parent / "stems")
|
||||
|
||||
client = AbletonSocketClient()
|
||||
clear_response = client.send("clear_project", {"keep_tracks": 0}, timeout=120.0)
|
||||
if clear_response.get("status") != "success":
|
||||
raise RuntimeError(clear_response.get("message", "No se pudo limpiar el proyecto"))
|
||||
|
||||
client.send("stop", {})
|
||||
client.send("set_tempo", {"tempo": round(profile["tempo"], 3)})
|
||||
client.send("show_arrangement_view", {})
|
||||
client.send("jump_to", {"time": 0})
|
||||
|
||||
created = []
|
||||
for (track_name, color, volume, muted), stem_key in zip(TRACK_LAYOUT, ("reference", "drums", "bass", "other", "vocals")):
|
||||
track_index = _create_track(client, track_name, color, volume)
|
||||
_import_full_length_audio(client, track_index, stems[stem_key], track_name)
|
||||
if muted:
|
||||
client.send("set_track_mute", {"index": track_index, "mute": True})
|
||||
created.append({
|
||||
"track_index": track_index,
|
||||
"name": track_name,
|
||||
"file_path": str(stems[stem_key]),
|
||||
})
|
||||
|
||||
_prepare_navigation_scenes(client, profile["structure"])
|
||||
client.send("loop_selection", {"start": 0, "length": max(32.0, round(profile["duration"] * profile["tempo"] / 60.0, 3)), "enable": False})
|
||||
client.send("jump_to", {"time": 0})
|
||||
client.send("show_arrangement_view", {})
|
||||
|
||||
session_info = client.send("get_session_info")
|
||||
return {
|
||||
"reference": str(reference_path),
|
||||
"tempo": profile["tempo"],
|
||||
"key": profile["key"],
|
||||
"structure": profile["structure"],
|
||||
"listener_device": profile["listener_device"],
|
||||
"stems": created,
|
||||
"session_info": session_info.get("result", {}),
|
||||
}
|
||||
|
||||
|
||||
def main() -> int:
|
||||
import argparse
|
||||
|
||||
parser = argparse.ArgumentParser(description="Rebuild an Ableton project directly from a reference track.")
|
||||
parser.add_argument("reference_path", help="Absolute or relative path to the reference audio file")
|
||||
args = parser.parse_args()
|
||||
|
||||
result = rebuild_project_from_reference(Path(args.reference_path))
|
||||
print(json.dumps(result, indent=2, ensure_ascii=False))
|
||||
return 0
|
||||
|
||||
|
||||
if __name__ == "__main__":
|
||||
raise SystemExit(main())
|
||||
Reference in New Issue
Block a user