modusa 0.4.19__tar.gz → 0.4.21__tar.gz

{modusa-0.4.19 → modusa-0.4.21}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.1
 Name: modusa
-Version: 0.4.19
+Version: 0.4.21
 Summary: A modular signal analysis python library.
 Author-Email: Ankit Anand <ankit0.anand0@gmail.com>
 License: MIT

{modusa-0.4.19 → modusa-0.4.21}/pyproject.toml

@@ -17,7 +17,7 @@ dependencies = [
 ]
 requires-python = ">=3.11"
 readme = "README.md"
-version = "0.4.19"
+version = "0.4.21"
 
 [project.license]
 text = "MIT"

{modusa-0.4.19 → modusa-0.4.21}/src/modusa/__init__.py

@@ -8,4 +8,4 @@ from modusa.tools import play, convert, record, save
 from modusa.tools import download
 from modusa.tools import load, load_ann
 
-__version__ = "0.4.19"
+__version__ = "0.4.21"

{modusa-0.4.19 → modusa-0.4.21}/src/modusa/tools/ann_loader.py

@@ -6,7 +6,7 @@
 # Email: ankit0.anand0@gmail.com
 #---------------------------------
 
-def load_ann(path, clip=None):
+def load_ann(path, trim=None):
 	"""
 	Load annotation from audatity label
 	text file and also ctm file.
@@ -15,9 +15,9 @@ def load_ann(path, clip=None):
 	----------
 	path: str
 		- label text/ctm file path.
-	clip: tuple[number, number] | number | None
-		- Incase you clipped the audio signal, this parameter will help clip the annotation.
-		- If you clip the audio, say from (10, 20), set the clip to (10, 20).
+	trim: tuple[number, number] | number | None
+		- Incase you trimmed the audio signal, this parameter will help clip the annotation making sure that the timings are aligned to the trimmed audio.
+		- If you trimmed the audio, say from (10, 20), set the trim to (10, 20).
 		- Default: None
 
 	Returns
@@ -42,14 +42,14 @@ def load_ann(path, clip=None):
 	ann = [] # This will store the annotation
 	
 	# Clipping the annotation to match with the clipped audio
-	if clip is not None:
+	if trim is not None:
 		# Map clip input to the right format
-		if isinstance(clip, int or float):
-			clip = (0, clip)
-		elif isinstance(clip, tuple) and len(clip) > 1:
-			clip = (clip[0], clip[1])
+		if isinstance(trim, int or float):
+			trim = (0, trim)
+		elif isinstance(trim, tuple) and len(trim) > 1:
+			trim = (trim[0], trim[1])
 		else:
-			raise ValueError(f"Invalid clip type or length: {type(clip)}, len={len(clip)}")
+			raise ValueError(f"Invalid clip type or length: {type(trim)}, len={len(trim)}")
 	
 	if path.suffix == ".txt":
 		with open(str(path), "r") as f:
@@ -60,11 +60,11 @@ def load_ann(path, clip=None):
 				
 				# Incase user has clipped the audio signal, we adjust the annotation
 				# to match the clipped audio
-				if clip is not None:
-					offset = clip[0]
+				if trim is not None:
+					offset = trim[0]
 					# Clamp annotation to clip boundaries
-					new_start = max(start, clip[0]) - offset
-					new_end   = min(end, clip[1]) - offset
+					new_start = max(start, trim[0]) - offset
+					new_end   = min(end, trim[1]) - offset
 					
 					# only keep if there's still overlap
 					if new_start < new_end:
@@ -89,11 +89,11 @@ def load_ann(path, clip=None):
 				
 				# Incase user has clipped the audio signal, we adjust the annotation
 				# to match the clipped audio
-				if clip is not None:
-					offset = clip[0]
+				if trim is not None:
+					offset = trim[0]
 					# Clamp annotation to clip boundaries
-					new_start = max(start, clip[0]) - offset
-					new_end   = min(end, clip[1]) - offset
+					new_start = max(start, trim[0]) - offset
+					new_end   = min(end, trim[1]) - offset
 					
 					# only keep if there's still overlap
 					if new_start < new_end:

{modusa-0.4.19 → modusa-0.4.21}/src/modusa/tools/audio_loader.py

@@ -3,14 +3,14 @@
 
 import soundfile as sf
 from scipy.signal import resample
+import numpy as np
 from pathlib import Path
 import tempfile
 from scipy.signal import resample
 from .youtube_downloader import download
 from .audio_converter import convert
 
-
-def load(path, sr=None, clip=None):
+def load(path, sr=None, trim=None, mono=True):
 	"""
 	Loads audio file from various sources.
 
@@ -19,19 +19,21 @@ def load(path, sr=None, clip=None):
 		import modusa as ms
 		audio_fp = ms.load(
 			"https://www.youtube.com/watch?v=lIpw9-Y_N0g",
-			sr = None, clip=(5, 10))
+			sr = None, trim=(5, 10))
 
 	Parameters
 	----------
 	path: str
-		- Path to the audio
-		- Youtube URL
+		- Path to the audio file.
+		- YouTube URL.
 	sr: int | None
 		- Sampling rate to load the audio in.
-	clip: number | tuple[number, number] | None
-		- Which segment of the audio you want.
-		- Eg., 10 => First 10 sec, (5, 10) => 5 to 10 second
+	trim: number | tuple[number, number] | None
+		- Segment of the audio to load.
+		- Example: 10 => First 10 seconds, (5, 10) => 5 to 10 seconds.
 		- Default: None => Entire audio.
+	mono: bool
+		- If True, loads the signal in mono.
 
 	Return
 	------
@@ -43,7 +45,6 @@ def load(path, sr=None, clip=None):
 		- Title of the loaded audio.
 		- Filename without extension or YouTube title.
 	"""
-	
 	# Check if the path is YouTube
 	if ".youtube." in str(path):
 		# Download the audio in temp directory using tempfile module
@@ -57,50 +58,52 @@ def load(path, sr=None, clip=None):
 			# Load the audio in memory
 			audio_data, audio_sr = sf.read(wav_audio_fp)
 			title = audio_fp.stem
-			
-			# Convert to mono if it's multi-channel
-			if audio_data.ndim > 1:
-				audio_data = audio_data.mean(axis=1)
-				
-			# Resample if needed
-			if sr is not None:
-				if audio_sr != sr:
-					n_samples = int(len(audio_data) * sr / audio_sr)
-					audio_data = resample(audio_data, n_samples)
-					audio_sr = sr
-		
 	else:
 		# Check if the file exists
 		fp = Path(path)
 		
 		if not fp.exists():
 			raise FileNotFoundError(f"{path} does not exist.")
-		
+			
 		# Load the audio in memory
 		audio_data, audio_sr = sf.read(fp)
 		title = fp.stem
 		
-		# Convert to mono if it's multi-channel
-		if audio_data.ndim > 1:
-			audio_data = audio_data.mean(axis=1)
+	# Convert to mono if requested and it's multi-channel
+	if mono and audio_data.ndim > 1:
+		audio_data = audio_data.mean(axis=1)
+		
+	# Resample if needed
+	if sr is not None and audio_sr != sr:
+		n_samples = int(len(audio_data) * sr / audio_sr)
+		
+		if audio_data.ndim == 1:
+			# Mono
+			audio_data = resample(audio_data, n_samples)
+		else:
+			# Stereo or multi-channel: resample each channel independently
+			audio_data = np.stack([
+				resample(audio_data[:, ch], n_samples)
+				for ch in range(audio_data.shape[1])
+			], axis=1)
 			
-		# Resample if needed
-		if sr is not None:
-			if audio_sr != sr:
-				n_samples = int(len(audio_data) * sr / audio_sr)
-				audio_data = resample(audio_data, n_samples)
-				audio_sr = sr
-				
-	# Clip the audio signal as per needed
-	if clip is not None:
-		# Map clip input to the right format
-		if isinstance(clip, int or float):
-			clip = (0, clip)
-		elif isinstance(clip, tuple) and len(clip) > 1:
-			clip = (clip[0], clip[1])
+		audio_sr = sr
+		
+	# Trim if requested
+	if trim is not None:
+		if isinstance(trim, (int, float)):
+			trim = (0, trim)
+		elif isinstance(trim, tuple) and len(trim) > 1:
+			trim = (trim[0], trim[1])
 		else:
-			raise ValueError(f"Invalid clip type or length: {type(clip)}, len={len(clip)}")
+			raise ValueError(f"Invalid trim type or length: {type(trim)}, len={len(trim)}")
+			
+		start = int(trim[0] * audio_sr)
+		end = int(trim[1] * audio_sr)
+		audio_data = audio_data[start:end]
+		
+	# Clip to avoid out-of-range playback issues
+	if np.issubdtype(audio_data.dtype, np.floating):
+		audio_data = np.clip(audio_data, -1.0, 1.0)
 		
-		audio_data = audio_data[int(clip[0]*audio_sr):int(clip[1]*audio_sr)]
-	
-	return audio_data, audio_sr, title
+	return audio_data.T, audio_sr, title

{modusa-0.4.19 → modusa-0.4.21}/src/modusa/tools/audio_player.py

@@ -1,28 +1,31 @@
 #!/usr/bin/env python3
 
-from IPython.display import Audio, HTML, display
-import numpy as np
-
-from IPython.display import Audio, HTML, display
-import numpy as np
-from pathlib import Path
-
 from IPython.display import Audio, HTML, display
 from pathlib import Path
 import numpy as np
 import base64
 
-def play(
-	y: np.ndarray,
-	sr: float,
-	clip: tuple[float, float] | None = None,
-	label: str | None = None,
-) -> None:
+def play(y, sr: float, clip=None, label=None):
 	"""
-	Audio player with optional clip selection, transcription-style label,
-	and an embedded bottom-right logo (../images/icon.png).
+	Audio player with optional clip selection, transcription-style label.
+
+	Parameters
+	----------
+	y: ndarray
+		- Audio signal.
+	sr: float
+		- Sampling rate.
+	clip: tuple[float, float] | None
+		- The portion from the audio signal to be played.
+	label: str | None
+		- Could be transcription/labels attached to the audio.
+	
+	Returns
+	-------
+	None
 	"""
-	start_time, end_time = 0.0, len(y) / sr
+	start_time = 0.0
+	end_time = len(y) / sr if y.ndim < 2 else y[0].size / sr
 	
 	# Optional clip selection
 	if clip is not None:
@@ -91,7 +94,7 @@ def play(
 		box-shadow:0 1px 3px rgba(0,0,0,0.05);
 	">
 		{label_html}
-		<div style="margin-top:10px;">
+		<div style="margin-top:10px; margin-bottom:10px">
 			{audio_html}
 		</div>
 		{logo_html}

{modusa-0.4.19 → modusa-0.4.21}/src/modusa/tools/plotter.py

@@ -209,6 +209,10 @@ class Fig:
 		
 		axs[-1, 0].tick_params(bottom=True, labelbottom=True)
 		
+		# Add the figure title on top-left (if any)
+		if self._fig_num is not None:
+			fig.suptitle(f'fig - {self._fig_num}', fontsize=12, fontweight='bold', x=0.01, ha='left', va='top', y=0.98)
+			
 		# xlim should be applied on reference subplot, rest all subplots will automatically adjust
 		if xlim is not None:
 			axs[0, 0].set_xlim(xlim)