idk whats going on man

components/AudioVisualizer.py

@@ -12,6 +12,7 @@ class AudioVisualizer(QtWidgets.QWidget):
         self.x_resolution = x_resolution
         self.fft_analyser = FFTAnalyser(self.media_player, self.x_resolution)
         self.fft_analyser.calculatedVisual.connect(self.set_amplitudes)
+        self.fft_analyser.calculatedVisualRs.connect(self.set_rs)
         self.fft_analyser.start()
         self.amps = np.array([])
         self._plot_item = None
@@ -55,6 +56,9 @@ class AudioVisualizer(QtWidgets.QWidget):
     def get_amplitudes(self):
         return self.amps
 
+    def get_rs(self):
+        return self.rs
+
     def get_decibels(self):
         """Convert amplitude values to decibel scale
 
@@ -63,7 +67,7 @@ class AudioVisualizer(QtWidgets.QWidget):
         With a noise floor cutoff at around -96dB (for very small values)
         """
         # Avoid log(0) by adding a small epsilon
-        epsilon = 1e-30
+        epsilon = 1e-6
         amplitudes = np.maximum(self.amps, epsilon)
         # Convert to decibels (20*log10 is the standard formula for amplitude to dB)
         db_values = 20 * np.log10(amplitudes)
@@ -71,6 +75,9 @@ class AudioVisualizer(QtWidgets.QWidget):
         db_values = np.maximum(db_values, -96)
         return db_values
 
+    def set_rs(self, rs):
+        self.rs = np.array(rs)
+
     def set_amplitudes(self, amps):
         """
         This function is hooked into the calculatedVisual signal from FFTAnalyzer() object

utils/fft_analyser.py

@@ -13,6 +13,7 @@ class FFTAnalyser(QtCore.QThread):
     """Analyses a song using FFTs."""
 
     calculatedVisual = QtCore.pyqtSignal(np.ndarray)
+    calculatedVisualRs = QtCore.pyqtSignal(np.ndarray)
 
     def __init__(self, player, x_resolution):  # noqa: F821
         super().__init__()
@@ -26,7 +27,7 @@ class FFTAnalyser(QtCore.QThread):
         # in this case, it takes 5% of the samples at some point in time
         self.sampling_window_length = 0.05
         self.visual_delta_threshold = 1000
-        self.sensitivity = 10
+        self.sensitivity = 1
 
     def reset_media(self):
         """Resets the media to the currently playing song."""
@@ -108,6 +109,7 @@ class FFTAnalyser(QtCore.QThread):
         # array (self.points) is so that we can fade out the previous amplitudes from
         # the past
         for n, amp in enumerate(point_samples):
+            amp *= 2
             if self.player.state() in (
                 self.player.PausedState,
                 self.player.StoppedState,
@@ -120,13 +122,14 @@ class FFTAnalyser(QtCore.QThread):
             else:
                 # Rise quickly to new peaks
                 self.points[n] = amp
-
+                # print(f'amp > points[n] - {amp} > {self.points[n]}')
             # Set a lower threshold to properly reach zero
-            if self.points[n] < 1e-4:
+            if self.points[n] < 1:
-                self.points[n] = 0
+                self.points[n] = 1e-5
 
+        # print(self.points)
         # interpolate points
-        rs = gaussian_filter1d(self.points, sigma=1)
+        rs = gaussian_filter1d(self.points, sigma=2)
 
         # divide by the highest sample in the song to normalise the
         # amps in terms of decimals from 0 -> 1

utils/fft_analyzer.py.ai

@@ -0,0 +1,117 @@
+# Credit
+# https://github.com/ravenkls/MilkPlayer/blob/master/audio/fft_analyser.py
+
+import time
+from PyQt5 import QtCore
+from pydub import AudioSegment
+import numpy as np
+from scipy.ndimage.filters import gaussian_filter1d
+from logging import debug, info
+
+
+class FFTAnalyser(QtCore.QThread):
+    """Analyses a song using FFTs."""
+
+    calculatedVisual = QtCore.pyqtSignal(np.ndarray)
+    calculatedVisualRs = QtCore.pyqtSignal(np.ndarray)
+
+    def __init__(self, player, x_resolution):  # noqa: F821
+        super().__init__()
+        self.player = player
+        self.reset_media()
+        self.player.currentMediaChanged.connect(self.reset_media)
+
+        self.resolution = x_resolution
+        # this length is a number, in seconds, of how much audio is sampled to determine the frequencies
+        # of the audio at a specific point in time
+        # in this case, it takes 5% of the samples at some point in time
+        self.sampling_window_length = 0.05
+        self.visual_delta_threshold = 1000
+        self.sensitivity = 0.2
+
+    def reset_media(self):
+        """Resets the media to the currently playing song."""
+        audio_file = self.player.currentMedia().canonicalUrl().path()
+        # if os.name == "nt" and audio_file.startswith("/"):
+        #     audio_file = audio_file[1:]
+        if audio_file:
+            try:
+                self.song = AudioSegment.from_file(audio_file).set_channels(1)
+            except PermissionError:
+                self.start_animate = False
+            else:
+                self.samples = np.array(self.song.get_array_of_samples())
+
+                self.max_sample = self.samples.max()
+                self.points = np.zeros(self.resolution)
+                self.start_animate = True
+        else:
+            self.start_animate = False
+
+    def calculate_amps(self):
+        """Calculates the amplitudes used for visualising the media."""
+
+        sample_count = int(self.song.frame_rate * self.sampling_window_length)
+        start_index = int((self.player.position() / 1000) * self.song.frame_rate)
+        # samples to analyse
+        v_sample = self.samples[start_index : start_index + sample_count]
+
+        # Use a window function to reduce spectral leakage
+        window = np.hanning(len(v_sample))
+        v_sample = v_sample * window
+
+        # use FFTs to analyse frequency and amplitudes
+        fourier = np.fft.fft(v_sample)
+        freq = np.fft.fftfreq(fourier.size, d=1/self.song.frame_rate)
+        amps = np.abs(fourier)[:len(fourier)//2]  # Only take positive frequencies
+        freq = freq[:len(fourier)//2]  # Match frequencies to amplitudes
+        
+        # Define frequency bands (in Hz)
+        bands = np.logspace(np.log10(10), np.log10(23000), self.resolution + 1)
+        point_samples = np.zeros(self.resolution)
+        
+        # Calculate average amplitude for each frequency band
+        for i in range(len(bands) - 1):
+            mask = (freq >= bands[i]) & (freq < bands[i+1])
+            if np.any(mask):
+                point_samples[i] = np.mean(amps[mask])
+        
+        # Calculate RMS of the sample for dynamic sensitivity
+        rms = np.sqrt(np.mean(np.square(v_sample)))
+        rms_ratio = min(0.2, rms / (0.01 * self.max_sample))  # Smooth transition near silence
+        
+        # Normalize and apply sensitivity with RMS-based scaling
+        if np.max(point_samples) > 0:
+            point_samples = point_samples / np.max(point_samples)
+            point_samples = point_samples * self.sensitivity * rms_ratio
+        else:
+            point_samples = np.zeros(self.resolution)
+
+        # Update visualization points with decay
+        for n in range(self.resolution):
+            amp = point_samples[n]
+            if self.player.state() in (self.player.PausedState, self.player.StoppedState):
+                self.points[n] *= 0.95  # Fast decay when paused/stopped
+            elif amp < self.points[n]:
+                # More aggressive decay for very quiet signals
+                decay_factor = 0.7 if rms_ratio < 0.1 else 0.9
+                self.points[n] = max(amp, self.points[n] * decay_factor)
+            else:
+                self.points[n] = amp
+
+        # Apply Gaussian smoothing
+        rs = gaussian_filter1d(self.points, sigma=1)
+        
+        # Emit the smoothed data
+        self.calculatedVisual.emit(rs)
+
+    def run(self):
+        """Runs the animate function depending on the song."""
+        while True:
+            if self.start_animate:
+                try:
+                    self.calculate_amps()
+                except ValueError:
+                    self.calculatedVisual.emit(np.zeros(self.resolution))
+                    self.start_animate = False
+            time.sleep(0.033)