From 42891006def34422ce3018f3e3d6d633a92ac654 Mon Sep 17 00:00:00 2001 From: translator Date: Thu, 5 Mar 2026 17:49:50 +0900 Subject: [PATCH] Added python services to log (debug or higher). Reworked live lyric preview to change custom background only for preview window and fixed mobile display. --- backend/job_manager.py | 61 +- backend/logging.yaml | 41 ++ backend/main.py | 10 + backend/requirements.txt | 5 +- backend/routers/projects.py | 8 +- backend/services/abc_generator.py | 37 +- backend/services/lrc_service.py | 8 +- backend/services/lyric_transcriber_service.py | 39 +- backend/services/mix_service.py | 7 + backend/services/pipeline_service.py | 9 + backend/services/separator_service.py | 10 + backend/services/transcriber_service.py | 6 + backend/static/index.html | 4 +- frontend/package-lock.json | 11 + frontend/package.json | 5 +- frontend/src/api/client.ts | 2 + .../src/components/AutoPipelineDialog.tsx | 1 + frontend/src/components/LyricsLivePreview.tsx | 56 +- frontend/src/pages/ProjectDetail.tsx | 641 +++++++++++++----- frontend/src/pages/Settings.tsx | 11 +- frontend/src/utils/autoPipeline.ts | 1 + start.sh | 2 +- 22 files changed, 756 insertions(+), 219 deletions(-) create mode 100644 backend/logging.yaml diff --git a/backend/job_manager.py b/backend/job_manager.py index 38045ca..69cd0d6 100755 --- a/backend/job_manager.py +++ b/backend/job_manager.py @@ -95,13 +95,24 @@ _flags_lock = threading.Lock() # --------------------------------------------------------------------------- # Thread pools # Two executors: -# _general_executor — multi-worker for CPU-bound tasks (e.g. ABC conversion) -# _gpu_executor — single-worker for GPU-exclusive tasks (PyTorch + TF) -# serialises Demucs/BS-RoFormer and Omnizart so they -# never compete for VRAM; FIFO by submission order. +# _general_executor — multi-worker for CPU-bound tasks (e.g. ABC conversion) +# _gpu_executor — single-worker for GPU-exclusive tasks (PyTorch + TF) +# serialises Demucs/BS-RoFormer and Omnizart so they +# never compete for VRAM; FIFO by submission order. +# _transcription_executor — single-worker queue ONLY for lyric transcription and +# LRC alignment. Kept separate from _gpu_executor so +# stem separation can run without blocking, but still +# prevents two transcription/alignment jobs from +# competing for VRAM. Jobs are submitted here instead +# of _gpu_executor for these job types. # --------------------------------------------------------------------------- -_general_executor = ThreadPoolExecutor(max_workers=4) -_gpu_executor = ThreadPoolExecutor(max_workers=1) +_general_executor = ThreadPoolExecutor(max_workers=4) +_gpu_executor = ThreadPoolExecutor(max_workers=1) +_transcription_executor = ThreadPoolExecutor(max_workers=1) + +# Queue-depth counter for the transcription executor (includes the running job). +_tx_queue_lock = threading.Lock() +_tx_queue_depth = 0 # --------------------------------------------------------------------------- @@ -207,14 +218,44 @@ def submit(fn: Callable, *args, **kwargs): def submit_gpu(fn: Callable, *args, **kwargs): """Submit a GPU-exclusive background task (single-worker FIFO queue). - Use for any job that loads PyTorch or TensorFlow models onto the GPU: - stem separation (Demucs / BS-RoFormer), MIDI transcription (Omnizart), - lyrics alignment (wav2vec2 + Demucs vocal extraction). Tasks run - strictly one-at-a-time in the order they were submitted. + Use for stem separation (Demucs / BS-RoFormer) and MIDI transcription. + Tasks run strictly one-at-a-time in the order they were submitted. """ _gpu_executor.submit(fn, *args, **kwargs) +def submit_transcription(job_id: str, fn: Callable, *args, **kwargs): + """Submit a lyric transcription or LRC alignment job. + + Uses a dedicated single-worker executor so transcription jobs are + serialised independently of stem separation, preventing VRAM contention. + Shows queue position in the job's progress field while waiting. + """ + global _tx_queue_depth + + with _tx_queue_lock: + _tx_queue_depth += 1 + position = _tx_queue_depth + + if position > 1: + update_job(job_id, progress=f"Queued (position {position} of {position})…") + + def _wrapper(): + global _tx_queue_depth + # Recalculate position at start time (earlier jobs may have finished). + with _tx_queue_lock: + pos = _tx_queue_depth + if pos > 1: + update_job(job_id, progress=f"Queued — waiting for {pos - 1} job(s) to finish…") + try: + fn(*args, **kwargs) + finally: + with _tx_queue_lock: + _tx_queue_depth -= 1 + + _transcription_executor.submit(_wrapper) + + def _record_to_dict(record: JobRecord) -> dict: return { "id": record.id, diff --git a/backend/logging.yaml b/backend/logging.yaml new file mode 100644 index 0000000..18debbb --- /dev/null +++ b/backend/logging.yaml @@ -0,0 +1,41 @@ +# StemEngine Python logging configuration. +# Edit `root.level` to control verbosity: +# DEBUG — full trace: job steps, model loads, GPU memory events, alignment details +# INFO — normal operational messages +# WARNING — warnings and errors only (quiet production mode) +# +# Individual loggers can be overridden in the `loggers` section below. + +version: 1 +disable_existing_loggers: false + +formatters: + standard: + format: "%(asctime)s %(levelname)-8s %(name)s — %(message)s" + datefmt: "%H:%M:%S" + +handlers: + console: + class: logging.StreamHandler + formatter: standard + stream: ext://sys.stdout + +root: + level: WARNING + handlers: [console] + +loggers: + # ── StemEngine application loggers ────────────────────────────────────────── + backend: + level: WARNING # change to DEBUG to see all app-level debug output + propagate: true + + # ── Third-party: kept at WARNING to avoid noise ─────────────────────────── + httpx: { level: WARNING, propagate: true } + httpcore: { level: WARNING, propagate: true } + urllib3: { level: WARNING, propagate: true } + multipart: { level: WARNING, propagate: true } + watchfiles: { level: WARNING, propagate: true } + asyncio: { level: WARNING, propagate: true } + h11: { level: WARNING, propagate: true } + uvicorn.access: { level: WARNING, propagate: true } diff --git a/backend/main.py b/backend/main.py index fc326bf..8a7396e 100755 --- a/backend/main.py +++ b/backend/main.py @@ -2,9 +2,19 @@ StemEngine — FastAPI application entry point. Mounts all API routers and serves the built React frontend as static files. """ +import logging +import logging.config from pathlib import Path from fastapi import FastAPI + +# Load logging config from the discrete YAML file next to this package. +# Edit backend/logging.yaml to change log levels without touching code. +_log_cfg = Path(__file__).parent / "logging.yaml" +if _log_cfg.exists(): + import yaml + with open(_log_cfg) as _f: + logging.config.dictConfig(yaml.safe_load(_f)) from fastapi.middleware.cors import CORSMiddleware from fastapi.staticfiles import StaticFiles from fastapi.responses import FileResponse diff --git a/backend/requirements.txt b/backend/requirements.txt index b295bd1..e29554a 100755 --- a/backend/requirements.txt +++ b/backend/requirements.txt @@ -51,9 +51,10 @@ cryptography>=42.0 # MIDI transcription via basic-pitch # Installed separately in Dockerfile with --no-deps to avoid TF/numpy conflict. # Runtime inference uses onnxruntime (above) instead of TensorFlow. -# pretty_midi is basic-pitch's MIDI output library; must be installed separately -# since it's normally pulled in as a dep of basic-pitch. +# pretty_midi and mir_eval are basic-pitch deps that must be installed separately +# since --no-deps skips them along with tensorflow. pretty_midi>=0.2.10 +mir_eval>=0.7 # Syllable hyphenation (pure-Python, for per-syllable ELRC timing) pyphen>=0.14.0 diff --git a/backend/routers/projects.py b/backend/routers/projects.py index 8929624..a3fea58 100755 --- a/backend/routers/projects.py +++ b/backend/routers/projects.py @@ -78,6 +78,7 @@ class LyricTranscribeRequest(BaseModel): diarize: bool = False # run speaker diarization (WhisperX only) lrc_power_preset: str = "standard" # 'eco' | 'standard' | 'performance' phoneme_align: bool = False # use phonemizer + phoneme model for syllable-level timestamps + two_pass: bool = False # run a second VAD pass to fill gaps > 2 s lyrics_source: str = "transcribe" # 'transcribe' | 'auto' | 'tag' | 'file' — when not 'transcribe', # skip Whisper and force-align existing lyrics instead existing_lyrics: Optional[str] = None # lyrics text (plain or LRC) to align; overrides lyrics_source @@ -407,7 +408,8 @@ def lrc_project(project_id: str, body: LrcRequest): job_id = jm.create_job("lrc_align") set_project_lrc_job(project_id, body.stem_name, job_id) - jm.submit_gpu( + jm.submit_transcription( + job_id, run_lrc_alignment, job_id, audio_path, @@ -481,7 +483,8 @@ def lyric_transcribe_project(project_id: str, body: LyricTranscribeRequest): # frontend always shows the most-recent AI transcription for that stem. set_project_lrc_job(project_id, f"{body.stem_name}_whisper", job_id) - jm.submit_gpu( + jm.submit_transcription( + job_id, run_lyric_transcription, job_id, stem_path, @@ -497,6 +500,7 @@ def lyric_transcribe_project(project_id: str, body: LyricTranscribeRequest): diarize=body.diarize, power_preset=body.lrc_power_preset, phoneme_align=body.phoneme_align, + two_pass=body.two_pass, existing_lyrics=resolved_lyrics, ) return {"job_id": job_id, "output_dir": str(output_dir)} diff --git a/backend/services/abc_generator.py b/backend/services/abc_generator.py index da7fc63..2484b8b 100755 --- a/backend/services/abc_generator.py +++ b/backend/services/abc_generator.py @@ -1,8 +1,12 @@ -from music21 import environment, converter, metadata, stream -from pathlib import Path -import subprocess -import os +import logging import traceback +import os +import subprocess +from pathlib import Path + +from music21 import environment, converter, metadata, stream + +log = logging.getLogger(__name__) def _post_process_abc_content(abc_text: str, title: str) -> str: @@ -47,7 +51,7 @@ def convert_midi_to_abc( for midi_path in midi_paths: part_name = Path(midi_path).stem part_title = f"{song_title} ({part_name})" - print(f" -> Processing part: {part_name}.mid") + log.debug(" -> Processing part: %s.mid", part_name) temp_midi_path = temp_dir / f"{part_title}.mid" temp_abc_path = temp_dir / f"{part_title}.abc" @@ -55,7 +59,7 @@ def convert_midi_to_abc( try: score_part = converter.parse(midi_path, forceSource=True) if not score_part or not score_part.flatten().notesAndRests: - print(f" -> Skipping empty MIDI: {Path(midi_path).name}") # noqa + log.debug(" -> Skipping empty MIDI: %s", Path(midi_path).name) continue single_part_score = stream.Score() @@ -71,12 +75,11 @@ def convert_midi_to_abc( result = subprocess.run(cmd, capture_output=True, text=True, check=False) if result.returncode != 0: - print(f" -> ERROR: midi2abc failed for {part_name}: " - f"{result.stderr}") + log.debug(" -> ERROR: midi2abc failed for %s: %s", part_name, result.stderr) continue if not temp_abc_path.exists() or temp_abc_path.stat().st_size == 0: - print(f" -> ERROR: midi2abc created an empty file for {part_name}.") + log.debug(" -> ERROR: midi2abc created an empty file for %s.", part_name) continue with open(temp_abc_path, 'r', encoding='utf-8') as f: @@ -84,12 +87,11 @@ def convert_midi_to_abc( abc_results[part_name] = _post_process_abc_content( abc_text_content, part_title) - print(f" -> Successfully converted {part_name} to ABC.") + log.debug(" -> Successfully converted %s to ABC.", part_name) except Exception as e: - print( - f" -> ERROR during ABC conversion for {part_name}: {e}") - traceback.print_exc() + log.debug(" -> ERROR during ABC conversion for %s: %s", part_name, e) + log.debug(traceback.format_exc()) finally: if temp_midi_path.exists(): os.remove(temp_midi_path) @@ -108,21 +110,20 @@ def generate_mix_abc(mix_midi_path: str, song_title: str) -> str | None: temp_abc_path = temp_dir / f"{mix_title}.abc" try: - print("\n -> Creating combined 'Mix' ABC score...") + log.debug(" -> Creating combined 'Mix' ABC score...") cmd = ['midi2abc', mix_midi_path, '-o', str(temp_abc_path)] result = subprocess.run(cmd, capture_output=True, text=True, check=False) if result.returncode == 0 and temp_abc_path.exists(): raw_mix_abc = temp_abc_path.read_text(encoding='utf-8') cleaned_mix_abc = _post_process_abc_content(raw_mix_abc, mix_title) - print(" -> Successfully created 'Mix' ABC.") + log.debug(" -> Successfully created 'Mix' ABC.") return cleaned_mix_abc else: - print( - f" -> ERROR: midi2abc failed for Mix: {result.stderr}") + log.debug(" -> ERROR: midi2abc failed for Mix: %s", result.stderr) return None except Exception as e: - print(f" -> ERROR during Mix ABC conversion: {e}") + log.debug(" -> ERROR during Mix ABC conversion: %s", e) return None finally: if temp_abc_path.exists(): diff --git a/backend/services/lrc_service.py b/backend/services/lrc_service.py index ae140ec..e6921c9 100755 --- a/backend/services/lrc_service.py +++ b/backend/services/lrc_service.py @@ -2,6 +2,7 @@ LRC alignment service for creating enhanced ELRC files with word-level timestamps. Uses the lyrics-sync module for vocal extraction and forced alignment. """ +import logging import os import re import sys @@ -12,6 +13,8 @@ from typing import Optional, List, Tuple from .. import job_manager as jm from ..config import JOBS_DIR, LRC_SYNC_DIR + +log = logging.getLogger(__name__) from .metadata_util import ( extract_lyrics_from_file, lyrics_to_plaintext, @@ -154,7 +157,9 @@ def run_lrc_alignment( else: out_path = LRC_SYNC_DIR / job_id out_path.mkdir(parents=True, exist_ok=True) - + log.debug("[%s] run_lrc_alignment start — source=%s vocals=%s precision=%s", + job_id, lyrics_source, is_vocals, timing_precision) + try: # Import lyrics-sync here to handle any import issues gracefully try: @@ -235,6 +240,7 @@ def run_lrc_alignment( ) except Exception as exc: + log.debug("[%s] lrc_alignment failed — %s: %s", job_id, type(exc).__name__, exc) jm.update_job(job_id, status=jm.JobStatus.ERROR, error=str(exc), progress="Error.") diff --git a/backend/services/lyric_transcriber_service.py b/backend/services/lyric_transcriber_service.py index 0cc1925..6424087 100755 --- a/backend/services/lyric_transcriber_service.py +++ b/backend/services/lyric_transcriber_service.py @@ -49,9 +49,13 @@ try: except Exception: pass +import logging + from .. import job_manager as jm from ..config import JOBS_DIR, LRC_SYNC_DIR +log = logging.getLogger(__name__) + @dataclass class WordTimestamp: @@ -1034,6 +1038,7 @@ def run_lyric_transcription( diarize: bool = False, power_preset: str = "standard", phoneme_align: bool = False, + two_pass: bool = False, existing_lyrics: Optional[str] = None, ): """ @@ -1064,7 +1069,9 @@ def run_lyric_transcription( else: out_path = LRC_SYNC_DIR / job_id out_path.mkdir(parents=True, exist_ok=True) - + log.debug("[%s] run_lyric_transcription start — model=%s engine=%s phoneme=%s two_pass=%s", + job_id, model, engine, phoneme_align, two_pass) + try: if engine == "whisperx": # ── WhisperX path: faster-whisper backend + phoneme-level alignment ── @@ -1087,11 +1094,13 @@ def run_lyric_transcription( return device = "cuda" if torch.cuda.is_available() else "cpu" + log.debug("[%s] device=%s power_preset=%s", job_id, device, power_preset) _wx_preset = _WHISPERX_POWER_PRESETS.get(power_preset, _WHISPERX_POWER_PRESETS['standard']) batch_size = _wx_preset['batch_size'] compute_type = _wx_preset['compute_type_cuda'] if device == "cuda" else "int8" # Auto-ensure vocals stem for best transcription quality. + log.debug("[%s] ensuring vocals stem from %s", job_id, audio_path) audio_path = _ensure_vocals(audio_path, job_id) # Music-tuned ASR options: @@ -1148,16 +1157,18 @@ def run_lyric_transcription( vad_options=_vad_opts, ) - jm.update_job(job_id, progress=f"Pass 1: Transcribing with WhisperX ({model}, {power_preset})...") + jm.update_job(job_id, progress=f"Transcribing with WhisperX ({model}, {power_preset})...") wx_result = wx_model.transcribe(audio_arr, batch_size=batch_size) detected_lang = wx_result.get("language") or language or "unknown" + log.debug("[%s] transcription done — lang=%s segments=%d", + job_id, detected_lang, len(wx_result.get("segments", []))) - # Pass 2: re-transcribe any gaps > 2 s that VAD may have missed - # (e.g. quieter verse sections after long instrumentals). - wx_result["segments"] = _fill_transcript_gaps( - wx_model, audio_arr, wx_result.get("segments", []), - batch_size, job_id, min_gap=2.0, - ) + if two_pass: + # Re-transcribe any gaps > 2 s that VAD may have missed + wx_result["segments"] = _fill_transcript_gaps( + wx_model, audio_arr, wx_result.get("segments", []), + batch_size, job_id, min_gap=2.0, + ) # Free the Whisper model from VRAM before loading the alignment model. # large-v2/v3 consume 3–6 GB; keeping both models resident simultaneously @@ -1173,6 +1184,7 @@ def run_lyric_transcription( return # ── Alignment (word-level or phoneme-level) ───────────────────────── + log.debug("[%s] starting alignment — phoneme=%s", job_id, phoneme_align) _use_phoneme = phoneme_align _phoneme_orig_words: list = [] @@ -1438,6 +1450,8 @@ def run_lyric_transcription( "word_count": sum(len(seg.words) for seg in result.segments), } + log.debug("[%s] complete — %d output files: %s", + job_id, len(output_files), [f["name"] for f in output_files]) jm.update_job( job_id, status=jm.JobStatus.DONE, @@ -1445,11 +1459,16 @@ def run_lyric_transcription( files=output_files, result=result_summary ) - + except Exception as exc: + log.debug("[%s] failed — %s: %s", job_id, type(exc).__name__, exc) jm.update_job(job_id, status=jm.JobStatus.ERROR, error=str(exc), progress="Error.") - # Best-effort GPU flush so a failed job doesn't strand VRAM. + # Best-effort GPU flush so a failed job doesn't strand VRAM for the next + # queued job. gc.collect() releases live tensor refs before empty_cache() + # can reclaim the CUDA memory they hold. try: + import gc + gc.collect() import torch if torch.cuda.is_available(): torch.cuda.empty_cache() diff --git a/backend/services/mix_service.py b/backend/services/mix_service.py index 6d14c51..8e906b2 100755 --- a/backend/services/mix_service.py +++ b/backend/services/mix_service.py @@ -2,12 +2,15 @@ Mix service: create static or dynamic mixes from separated stems. CPU-bound (pydub) — uses the general executor, not the GPU queue. """ +import logging from pathlib import Path from typing import Dict, List, Optional from .. import job_manager as jm from .preset_service import make_static, make_dynamic +log = logging.getLogger(__name__) + def run_mix( job_id: str, @@ -19,6 +22,8 @@ def run_mix( dynamic_levels: Optional[Dict[str, float]] = None, ): """Background task: create a mix from separated stems.""" + log.debug("[%s] run_mix start — preset=%s format=%s output=%s", + job_id, preset_type, output_format, output_path) jm.update_job(job_id, status=jm.JobStatus.RUNNING, progress="Creating mix...") try: out = Path(output_path) @@ -33,6 +38,7 @@ def run_mix( else: raise ValueError(f"Unknown preset type: {preset_type}") + log.debug("[%s] mix complete — output=%s", job_id, result_path) jm.update_job( job_id, status=jm.JobStatus.DONE, @@ -40,6 +46,7 @@ def run_mix( files=[{"name": out.name, "path": result_path, "type": preset_type}], ) except Exception as exc: + log.debug("[%s] mix failed — %s: %s", job_id, type(exc).__name__, exc) jm.update_job(job_id, status=jm.JobStatus.ERROR, error=str(exc), progress="Error.") diff --git a/backend/services/pipeline_service.py b/backend/services/pipeline_service.py index b079ce4..795fa4b 100755 --- a/backend/services/pipeline_service.py +++ b/backend/services/pipeline_service.py @@ -3,6 +3,7 @@ Full pipeline: separate → transcribe → abc-convert. Supports both Demucs (CLI) and BS-RoFormer (Python API) for the separation stage. Uses project-based folder structure for organized output. """ +import logging import sys import subprocess import time @@ -16,6 +17,8 @@ from .project_namer import project_name, make_project_dirs from .separator_service import BSR_PREFIX, BSR_STEM_MODE from . import abc_generator +log = logging.getLogger(__name__) + def run_pipeline( job_id: str, @@ -34,6 +37,9 @@ def run_pipeline( Creates organized project folder structure under MEDIA_DIR. """ + log.debug("[%s] run_pipeline start — model=%s format=%s transcribe_mode=%s input=%s", + job_id, sep_model, sep_format, transcribe_mode, input_path) + # Read metadata for project naming jm.update_job(job_id, status=jm.JobStatus.RUNNING, progress="Reading metadata...") metadata = read_metadata(input_path) @@ -78,6 +84,7 @@ def run_pipeline( job_id, input_path, sep_model, sep_format, stems_dir ) except Exception as exc: + log.debug("[%s] pipeline separation failed — %s: %s", job_id, type(exc).__name__, exc) jm.update_job(job_id, status=jm.JobStatus.ERROR, error=str(exc), progress="Separation failed.") return @@ -152,6 +159,8 @@ def run_pipeline( except Exception as exc: jm.update_job(job_id, progress=f"[3/3] ABC conversion error: {exc}") + log.debug("[%s] pipeline complete — project=%s stems=%d files=%d", + job_id, proj_name, len(stem_files), len(all_files)) jm.update_job( job_id, status=jm.JobStatus.DONE, progress="Pipeline complete.", diff --git a/backend/services/separator_service.py b/backend/services/separator_service.py index 2db1796..d43dba3 100755 --- a/backend/services/separator_service.py +++ b/backend/services/separator_service.py @@ -4,6 +4,7 @@ Stem separation service with project-based folder structure and preset mixing. - BS-RoFormer → Python API (in-process, no subprocess needed) - Presets → Instrumental, Static, Dynamic mixing via preset_service """ +import logging import os import subprocess import sys @@ -24,6 +25,8 @@ from .project_namer import project_name, make_project_dirs from .preset_service import make_instrumental, make_static, make_dynamic from .project_metadata import write_project_metadata +log = logging.getLogger(__name__) + # --------------------------------------------------------------------------- # Stem file naming helper # --------------------------------------------------------------------------- @@ -229,6 +232,9 @@ def run_separation( Static/ (if preset is static) Dynamic/ (if preset is dynamic) """ + log.debug("[%s] run_separation start — model=%s format=%s preset=%s input=%s", + job_id, model, output_format, preset_type, input_path) + # Read metadata for project naming jm.update_job(job_id, status=jm.JobStatus.RUNNING, progress="Reading metadata...") metadata = read_metadata(input_path) @@ -565,6 +571,8 @@ def run_separation( jm.append_job_files(parent_job_id, stem_entries) project_id_result = None + log.debug("[%s] separation complete — project=%s stems=%d files=%d", + job_id, proj_name, len(stem_files), len(all_files)) jm.update_job( job_id, status=jm.JobStatus.DONE, @@ -600,8 +608,10 @@ def run_separation( "The GPU was automatically reset — please retry the separation. " "If the error keeps occurring, restart the container." ) + log.debug("[%s] separation GPU error — %s: %s", job_id, type(exc).__name__, err_str[:200]) jm.update_job(job_id, status=jm.JobStatus.ERROR, error=user_msg, progress="GPU error — retry to use reset GPU.") else: + log.debug("[%s] separation failed — %s: %s", job_id, type(exc).__name__, exc) jm.update_job(job_id, status=jm.JobStatus.ERROR, error=err_str, progress="Error.") diff --git a/backend/services/transcriber_service.py b/backend/services/transcriber_service.py index 61e9cdf..4224a98 100755 --- a/backend/services/transcriber_service.py +++ b/backend/services/transcriber_service.py @@ -4,11 +4,14 @@ Supports two modes mapped to inference parameter presets: polyphonic — general multi-instrument transcription melody — single-voice / vocal melody (melodia trick, min frequency) """ +import logging from pathlib import Path from .. import job_manager as jm from ..config import JOBS_DIR, TRANSCRIBE_PRESETS +log = logging.getLogger(__name__) + def run_transcription(job_id: str, input_path: str, mode: str): """ @@ -18,6 +21,7 @@ def run_transcription(job_id: str, input_path: str, mode: str): from basic_pitch.inference import predict from basic_pitch import ICASSP_2022_MODEL_PATH + log.debug("[%s] run_transcription start — mode=%s input=%s", job_id, mode, input_path) output_dir = JOBS_DIR / job_id output_dir.mkdir(parents=True, exist_ok=True) @@ -45,6 +49,7 @@ def run_transcription(job_id: str, input_path: str, mode: str): dest = output_dir / output_name midi_data.write(str(dest)) + log.debug("[%s] transcription complete — output=%s", job_id, dest.name) jm.update_job( job_id, status=jm.JobStatus.DONE, @@ -53,4 +58,5 @@ def run_transcription(job_id: str, input_path: str, mode: str): ) except Exception as exc: + log.debug("[%s] transcription failed — %s: %s", job_id, type(exc).__name__, exc) jm.update_job(job_id, status=jm.JobStatus.ERROR, error=str(exc), progress="Error.") diff --git a/backend/static/index.html b/backend/static/index.html index 69610ca..7a2db23 100644 --- a/backend/static/index.html +++ b/backend/static/index.html @@ -19,8 +19,8 @@ - - + +
diff --git a/frontend/package-lock.json b/frontend/package-lock.json index 82da8f3..ead39f9 100755 --- a/frontend/package-lock.json +++ b/frontend/package-lock.json @@ -10,6 +10,7 @@ "dependencies": { "@tanstack/react-query": "^5.0.0", "@tonejs/midi": "^2.0.28", + "abcjs": "^6.6.2", "lucide-react": "^0.294.0", "react": "^18.2.0", "react-dom": "^18.2.0", @@ -1214,6 +1215,16 @@ "vite": "^4.2.0 || ^5.0.0 || ^6.0.0 || ^7.0.0" } }, + "node_modules/abcjs": { + "version": "6.6.2", + "resolved": "https://registry.npmjs.org/abcjs/-/abcjs-6.6.2.tgz", + "integrity": "sha512-YLbp5lYUq0uOywWZx9EuTdm0TcflKZi7hOzz366A/LFl3qoAXSYIjznJQmr/VeHg8NcLxZYoN8dLi7PqCpxKEA==", + "license": "MIT", + "funding": { + "type": "github", + "url": "https://github.com/sponsors/paulrosen" + } + }, "node_modules/any-promise": { "version": "1.3.0", "resolved": "https://registry.npmjs.org/any-promise/-/any-promise-1.3.0.tgz", diff --git a/frontend/package.json b/frontend/package.json index 4234cc5..547ca61 100755 --- a/frontend/package.json +++ b/frontend/package.json @@ -11,10 +11,11 @@ "dependencies": { "@tanstack/react-query": "^5.0.0", "@tonejs/midi": "^2.0.28", - "tone": "^14.7.0", + "abcjs": "^6.6.2", "lucide-react": "^0.294.0", "react": "^18.2.0", - "react-dom": "^18.2.0" + "react-dom": "^18.2.0", + "tone": "^14.7.0" }, "devDependencies": { "@types/react": "^18.2.37", diff --git a/frontend/src/api/client.ts b/frontend/src/api/client.ts index efcc181..c07b789 100755 --- a/frontend/src/api/client.ts +++ b/frontend/src/api/client.ts @@ -389,6 +389,7 @@ export const lyricTranscribeProject = ( diarize?: boolean lrcPowerPreset?: string phonemeAlign?: boolean + twoPass?: boolean } ) => { const wordTs = opts?.wordTimestamps ?? true @@ -414,6 +415,7 @@ export const lyricTranscribeProject = ( diarize: opts?.diarize ?? false, lrc_power_preset: opts?.lrcPowerPreset ?? 'standard', phoneme_align: phonemeAlign, + two_pass: opts?.twoPass ?? false, }), }) } diff --git a/frontend/src/components/AutoPipelineDialog.tsx b/frontend/src/components/AutoPipelineDialog.tsx index d6584b1..6e3b4ee 100755 --- a/frontend/src/components/AutoPipelineDialog.tsx +++ b/frontend/src/components/AutoPipelineDialog.tsx @@ -185,6 +185,7 @@ export default function AutoPipelineDialog({ target, onClose, onComplete, onOpen wordTimestamps: wordTs, diarize: step.diarize, phonemeAlign: step.phonemeAlign, + twoPass: step.twoPass ?? false, }) await pollUntilDone(transcribeJobId, ctrl.signal) }, ctrl.signal, retry) diff --git a/frontend/src/components/LyricsLivePreview.tsx b/frontend/src/components/LyricsLivePreview.tsx index 0aabb88..90f5da7 100755 --- a/frontend/src/components/LyricsLivePreview.tsx +++ b/frontend/src/components/LyricsLivePreview.tsx @@ -843,6 +843,7 @@ export default function LyricsLivePreview({ lrcText, audioUrl, stems, title, sty // ── Fullscreen ──────────────────────────────────────────────────────────── const fullscreenRef = useRef(null) const [isFullscreen, setIsFullscreen] = useState(false) + const [isNarrow, setIsNarrow] = useState(typeof window !== 'undefined' && window.innerWidth < 640) // Scale font size up in fullscreen so lyrics fill the larger canvas const fontSize = isFullscreen ? Math.round(baseFontSize * 1.65) : baseFontSize @@ -853,6 +854,12 @@ export default function LyricsLivePreview({ lrcText, audioUrl, stems, title, sty return () => document.removeEventListener('fullscreenchange', onFsChange) }, []) + useEffect(() => { + function check() { setIsNarrow(window.innerWidth < 640) } + window.addEventListener('resize', check) + return () => window.removeEventListener('resize', check) + }, []) + function toggleFullscreen() { if (!isFullscreen) fullscreenRef.current?.requestFullscreen() else document.exitFullscreen() @@ -1537,8 +1544,7 @@ export default function LyricsLivePreview({ lrcText, audioUrl, stems, title, sty function renderTransport() { return (
{open && ( - err - ?

{err}

- : !midi - ?
- : <> - + loading + ?
+ : err + ?

{err}

+ : midiList.length > 0 + ? <> + - {/* Playback transport bar */} -
- - {sfState === 'ready' && ( - - )} - - {sfState === 'idle' && 'Grand Piano · click Play to load'} - {sfState === 'loading' && 'Fetching piano samples…'} - {sfState === 'ready' && 'Grand Piano'} - {sfState === 'error' && 'Soundfont load failed'} - -
- +
+ {/* Seek bar */} +
+ {curStr} + handleSeek(parseFloat(e.target.value))} + disabled={sfState !== 'ready'} + className="flex-1 h-1.5 bg-gray-700 rounded-lg appearance-none cursor-pointer accent-purple-500 disabled:opacity-40" + /> + {durStr} +
+ + {/* Play / Pause / Stop */} +
+ + {sfState === 'ready' && ( + + )} + + {sfState === 'idle' && `${totalNotes} notes · click Play`} + {sfState === 'loading' && 'Loading instruments…'} + {sfState === 'ready' && `${totalNotes} notes`} + {sfState === 'error' && 'Failed to load'} + +
+ + {/* Per-stem rows: M / S / label / instrument / download */} +
+ {stems.map((stem, si) => { + const program = stemInstruments[si] ?? 0 + const isMuted = mutedStems[si] ?? false + const isSoloed = soloedStems[si] ?? false + const isHidden = hiddenStems[si] ?? false + return ( +
+ + + + {stem.label} + + e.stopPropagation()} + title={`Download ${stem.file.name}`} + className="text-gray-600 hover:text-purple-400 transition-colors flex-shrink-0" + > + + +
+ ) + })} +
+
+ + : null )}
) @@ -857,13 +1129,6 @@ function StemMidiRow({ {transJob && isActive(transJob) && (

{transJob.progress}

)} - {transJob?.status === 'done' && ( -
- {transJob.files.filter((f) => f.name.endsWith('.mid')).map((f) => ( - - ))} -
- )} {transJob?.status === 'error' && (

{transJob.error}

)} @@ -951,6 +1216,21 @@ function MidiSection({ })} )} + {(() => { + const doneStems: MidiGroupStem[] = stems + .map(stem => { + const stemName = stem.stem ?? stem.name + const job = transJobs[stemName] + if (job?.status !== 'done') return null + const midFile = job.files.find(f => f.name.endsWith('.mid')) + if (!midFile) return null + const label = stemName.charAt(0).toUpperCase() + stemName.slice(1) + return { label, file: midFile, jobId: job.id } + }) + .filter((x): x is MidiGroupStem => x !== null) + if (!doneStems.length) return null + return s.jobId).join(',')} stems={doneStems} /> + })()} )} ) @@ -1141,6 +1421,8 @@ function AbcFileViewer({ }: { file: JobFile; jobId: string; expanded: boolean; onToggle: () => void }) { const [text, setText] = useState(null) const [loading, setLoading] = useState(false) + const [view, setView] = useState<'sheet' | 'text'>('sheet') + const svgRef = useRef(null) async function load() { if (text !== null) { onToggle(); return } @@ -1153,21 +1435,67 @@ function AbcFileViewer({ onToggle() } + // Render / re-render sheet music whenever visible + useEffect(() => { + if (!expanded || view !== 'sheet' || !text || !svgRef.current) return + import('abcjs').then(({ renderAbc }) => { + renderAbc(svgRef.current!, text, { + responsive: 'resize', + add_classes: true, + paddingbottom: 20, + paddingtop: 16, + paddingright: 16, + paddingleft: 16, + wrap: { minSpacing: 1.8, maxSpacing: 2.7, preferredMeasuresPerLine: 4 }, + }) + }).catch(() => {}) + }, [expanded, view, text]) + return (
+ {/* Header */}
- - - - +
+ {expanded && text !== null && ( +
+ + +
+ )} + + + +
+ {expanded && text !== null && ( -
-          {text}
-        
+ view === 'sheet' + ?
+ :
+              {text}
+            
)}
) @@ -1348,6 +1676,7 @@ function StemLyricsRow({ onAction: () => void }) { const [mode, setMode] = useState<'align' | 'transcribe'>('transcribe') + const [twoPass, setTwoPass] = useState(false) // activeJob is the job relevant to the current mode tab const activeJob = mode === 'transcribe' ? whisperJob : lrcJob // Track job IDs started in this session — error state is only shown for fresh jobs @@ -1493,6 +1822,7 @@ function StemLyricsRow({ diarize, lrcPowerPreset: powerPreset, phonemeAlign, + twoPass, }) sessionJobIds.current.add(res.job_id) onAction() @@ -1517,11 +1847,11 @@ function StemLyricsRow({ {label}
- +
{activeJob && (activeJob.status !== 'error' || showError) && } @@ -1627,6 +1957,15 @@ function StemLyricsRow({ /> Phonemes +