Upload main.py

main.py

@@ -178,10 +178,6 @@ async def transcribe(
         description="Highest frequency to transcribe in Hz (e.g. 2000 for voice). "
                     "Leave empty for no upper limit.",
     ),
-    include_pitch_bends: bool = Form(
-        True,
-        description="Whether to detect and return sub-semitone pitch bend data.",
-    ),
     multiple_pitch_bends: bool = Form(
         False,
         description="Allow multiple simultaneous pitch bends (polyphonic pitch bend). "
@@ -231,24 +227,34 @@ async def transcribe(
         tmp.write(audio_bytes)
         tmp_path = tmp.name
 
-    try:
-        # ── Validate & clamp user settings ───────────────────────────────────
-        onset_threshold = _clamp(onset_threshold, 0.05, 0.95)
-        frame_threshold = _clamp(frame_threshold, 0.05, 0.95)
-        min_note_length = max(0.01, min_note_length)
+    # ── Validate & clamp user settings (before try so always defined) ────────
+    onset_threshold = _clamp(onset_threshold, 0.05, 0.95)
+    frame_threshold = _clamp(frame_threshold, 0.05, 0.95)
+    min_note_length = max(0.01, min_note_length)
+    min_note_length_ms = min_note_length * 1000.0  # basic-pitch wants ms, not s
 
+    try:
         # ── Run basic-pitch ───────────────────────────────────────────────────
+        # Exact signature (basic-pitch >= 0.3.x):
+        #   predict(audio_path, model_or_model_path,
+        #           onset_threshold, frame_threshold,
+        #           minimum_note_length,   ← in MILLISECONDS, not seconds
+        #           minimum_frequency, maximum_frequency,
+        #           multiple_pitch_bends, melodia_trick,
+        #           debug_file, midi_tempo)
+        # NOTE: `include_pitch_bends` was removed from predict(); pitch bends
+        # are now always detected internally and stored in the MIDI output.
+
         _model_output, midi_data, note_events = predict(
             tmp_path,
-            _MODEL,                         # pre-loaded — no cold start per request
-            onset_threshold=onset_threshold,
-            frame_threshold=frame_threshold,
-            minimum_note_length=min_note_length,
-            minimum_frequency=min_frequency,
-            maximum_frequency=max_frequency,
-            include_pitch_bends=include_pitch_bends,
-            multiple_pitch_bends=multiple_pitch_bends,
-            melodia_trick=melodia_trick,
+            _MODEL,                 # pre-loaded — no cold start per request
+            onset_threshold,
+            frame_threshold,
+            min_note_length_ms,
+            min_frequency,          # None = no lower bound
+            max_frequency,          # None = no upper bound
+            multiple_pitch_bends,
+            melodia_trick,
         )
 
     except Exception as exc:
@@ -277,10 +283,10 @@ async def transcribe(
         settings={
             "onset_threshold": onset_threshold,
             "frame_threshold": frame_threshold,
-            "min_note_length": min_note_length,
+            "min_note_length_seconds": min_note_length,
+            "min_note_length_ms": min_note_length_ms,
             "min_frequency": min_frequency,
             "max_frequency": max_frequency,
-            "include_pitch_bends": include_pitch_bends,
             "multiple_pitch_bends": multiple_pitch_bends,
             "melodia_trick": melodia_trick,
         },