feat: add music sync viz modes and auto_gain audio filter

Add 4 new audio visualization modes powered by MusicAnalyzer:
- pulse_on_beat: BPM-synced pulsing with smooth beat phase
- energy_gradient: bass/mid/treble mapped to scrolling gradient
- spectrum_bands: three VU zones for frequency bands
- strobe_on_drop: state-driven strobe on detected musical drops

MusicAnalyzer provides BPM estimation (median IBI), beat phase tracking,
asymmetric energy envelope, 3-band frequency splitting, and drop
detection state machine (idle/buildup/drop/recovery).

Add auto_gain audio filter for automatic level normalization via rolling
peak tracking with configurable target level and response time.

Deprecate auto_gain on Audio Value Source (use the filter instead).

server/src/wled_controller/api/routes/color_strip_sources.py

@@ -191,6 +191,7 @@ _RESPONSE_MAP: dict = {
         color=s.color.to_dict(),
         color_peak=s.color_peak.to_dict(),
         mirror=s.mirror,
+        beat_decay=s.beat_decay.to_dict(),
     ),
     ApiInputColorStripSource: lambda s, kw: ApiInputCSSResponse(
         **kw,

server/src/wled_controller/api/schemas/color_strip_sources.py

@@ -164,6 +164,7 @@ class AudioCSSResponse(_CSSResponseBase):
     color: Any = Field(description="Primary color")
     color_peak: Any = Field(description="Peak color")
     mirror: bool = Field(description="Mirror mode")
+    beat_decay: Any = Field(default=0.15, description="Beat pulse decay rate (music modes)")
 
 
 class ApiInputCSSResponse(_CSSResponseBase):
@@ -340,6 +341,9 @@ class AudioCSSCreate(_CSSCreateBase):
     color: Any = Field(default=None, description="Primary color")
     color_peak: Any = Field(default=None, description="Peak color [R,G,B]")
     mirror: Optional[bool] = Field(None, description="Mirror mode")
+    beat_decay: Any = Field(
+        default=None, description="Beat pulse decay rate (music modes, 0.01-0.5)"
+    )
 
 
 class ApiInputCSSCreate(_CSSCreateBase):
@@ -527,6 +531,7 @@ class AudioCSSUpdate(_CSSUpdateBase):
     color: Any = Field(default=None, description="Primary color")
     color_peak: Any = Field(default=None, description="Peak color [R,G,B]")
     mirror: Optional[bool] = Field(None, description="Mirror mode")
+    beat_decay: Any = Field(default=None, description="Beat pulse decay rate (music modes)")
 
 
 class ApiInputCSSUpdate(_CSSUpdateBase):

server/src/wled_controller/core/audio/filters/__init__.py

@@ -21,6 +21,7 @@ import wled_controller.core.audio.filters.compressor  # noqa: F401
 import wled_controller.core.audio.filters.inverter  # noqa: F401
 import wled_controller.core.audio.filters.beat_gate  # noqa: F401
 import wled_controller.core.audio.filters.delay  # noqa: F401
+import wled_controller.core.audio.filters.auto_gain  # noqa: F401
 
 __all__ = [
     "AudioFilter",

server/src/wled_controller/core/audio/filters/auto_gain.py

@@ -0,0 +1,84 @@
+"""Auto-gain audio filter — automatic level normalization.
+
+Tracks a rolling peak of the audio signal and scales all levels so the
+output uses the full 0-1 range regardless of input volume.  Zero-config
+by default; optional parameters for tuning response speed and target level.
+"""
+
+from dataclasses import replace
+from typing import Any, Dict, List
+
+import numpy as np
+
+from wled_controller.core.audio.analysis import AudioAnalysis
+from wled_controller.core.audio.filters.base import AudioFilter, AudioFilterOptionDef
+from wled_controller.core.audio.filters.registry import AudioFilterRegistry
+
+
+@AudioFilterRegistry.register
+class AutoGainFilter(AudioFilter):
+    """Normalize audio levels against a rolling observed peak.
+
+    Tracks the maximum signal level over a configurable time window and
+    scales all values so the loudest recent signal maps to ``target_level``.
+    The rolling peak decays slowly so the gain adapts to changing volumes.
+    """
+
+    filter_id = "auto_gain"
+    filter_name = "Auto Gain"
+
+    def __init__(self, options: Dict[str, Any]):
+        super().__init__(options)
+        self._target = self.options["target_level"]
+        # Decay factor per frame (~30 fps): controls how fast the peak forgets.
+        # response_time seconds → peak halves in that time.
+        response_time = self.options["response_time"]
+        # decay = 0.5^(1 / (fps * response_time))  ≈ e^(-ln2 / (fps * t))
+        self._decay = 0.5 ** (1.0 / max(1.0, 30.0 * response_time))
+        self._rolling_peak = 0.0
+
+    @classmethod
+    def get_options_schema(cls) -> List[AudioFilterOptionDef]:
+        return [
+            AudioFilterOptionDef(
+                key="target_level",
+                label="Target Level",
+                option_type="float",
+                default=0.8,
+                min_value=0.1,
+                max_value=1.0,
+                step=0.05,
+            ),
+            AudioFilterOptionDef(
+                key="response_time",
+                label="Response Time (s)",
+                option_type="float",
+                default=3.0,
+                min_value=0.5,
+                max_value=15.0,
+                step=0.5,
+            ),
+        ]
+
+    def process(self, analysis: AudioAnalysis) -> AudioAnalysis:
+        # Track the rolling peak from the loudest signal component
+        current_peak = max(analysis.rms, analysis.peak)
+        self._rolling_peak = max(current_peak, self._rolling_peak * self._decay)
+
+        if self._rolling_peak < 0.001:
+            return analysis  # signal too quiet, don't amplify noise
+
+        factor = self._target / self._rolling_peak
+        if 0.95 <= factor <= 1.05:
+            return analysis  # close enough, avoid jitter
+
+        return replace(
+            analysis,
+            rms=min(1.0, analysis.rms * factor),
+            peak=min(1.0, analysis.peak * factor),
+            spectrum=np.clip(analysis.spectrum * factor, 0.0, 1.0).astype(np.float32),
+            left_rms=min(1.0, analysis.left_rms * factor),
+            left_spectrum=np.clip(analysis.left_spectrum * factor, 0.0, 1.0).astype(np.float32),
+            right_rms=min(1.0, analysis.right_rms * factor),
+            right_spectrum=np.clip(analysis.right_spectrum * factor, 0.0, 1.0).astype(np.float32),
+        )

server/src/wled_controller/core/audio/music_analyzer.py

@@ -0,0 +1,223 @@
+"""Stateful music feature extractor — BPM, beat phase, energy, drop detection.
+
+Consumes AudioAnalysis snapshots and produces higher-level MusicFeatures
+for semantic music-reactive visualizations.  Pure numpy, no external deps.
+"""
+
+import time
+from dataclasses import dataclass
+
+import numpy as np
+
+# Frequency band boundaries (indices into 64-band log spectrum)
+_BASS_END = 10  # bins 0-9  ≈ 20-250 Hz
+_MID_END = 40  # bins 10-39 ≈ 250-4000 Hz
+# bins 40-63 ≈ 4000-20000 Hz (treble)
+
+_MAX_BEAT_HISTORY = 30
+_BPM_MIN = 40.0
+_BPM_MAX = 220.0
+
+# Drop detection timing
+_BUILDUP_MIN_DURATION = 1.0  # seconds of rising energy before "buildup"
+_DROP_THRESHOLD = 0.5  # energy must drop by this fraction
+_DROP_RECOVERY_TIME = 0.5  # seconds in "drop" state before "recovery"
+_RECOVERY_DURATION = 1.0  # seconds in "recovery" before returning to idle
+_MIN_ENERGY_FOR_BUILDUP = 0.15  # minimum energy to enter buildup
+
+
+@dataclass(frozen=True)
+class MusicFeatures:
+    """Snapshot of semantic music features for one frame."""
+
+    bpm: float = 0.0
+    beat: bool = False
+    beat_intensity: float = 0.0
+    beat_phase: float = 0.0  # 0.0-1.0 position in current beat cycle
+    energy: float = 0.0  # smoothed RMS 0.0-1.0
+    energy_delta: float = 0.0  # rate of change
+    bass_energy: float = 0.0
+    mid_energy: float = 0.0
+    treble_energy: float = 0.0
+    drop_state: str = "idle"  # idle | buildup | drop | recovery
+    drop_intensity: float = 0.0
+
+
+class MusicAnalyzer:
+    """Extracts semantic music features from AudioAnalysis snapshots.
+
+    Owned by a single stream thread — not thread-safe, no locking needed.
+    """
+
+    def __init__(self, sensitivity: float = 1.0):
+        self._sensitivity = sensitivity
+
+        # Beat history (timestamps of detected beats)
+        self._beat_times: list[float] = []
+        self._bpm: float = 0.0
+        self._last_beat_time: float = 0.0
+
+        # Energy tracking
+        self._smoothed_energy: float = 0.0
+        self._prev_energy: float = 0.0
+        self._energy_attack = 0.15  # fast attack
+        self._energy_release = 0.05  # slow release
+
+        # Band energy smoothing
+        self._bass: float = 0.0
+        self._mid: float = 0.0
+        self._treble: float = 0.0
+        self._band_smoothing = 0.25
+
+        # Drop detection state machine
+        self._drop_state = "idle"
+        self._drop_state_time: float = 0.0  # when current state started
+        self._buildup_start_energy: float = 0.0
+        self._peak_energy: float = 0.0
+        self._drop_intensity: float = 0.0
+
+    @property
+    def sensitivity(self) -> float:
+        return self._sensitivity
+
+    @sensitivity.setter
+    def sensitivity(self, value: float) -> None:
+        self._sensitivity = max(0.1, value)
+
+    def update(self, analysis) -> MusicFeatures:
+        """Process one AudioAnalysis frame and return MusicFeatures."""
+        now = time.perf_counter()
+        spectrum = analysis.spectrum
+        rms = analysis.rms * self._sensitivity
+        beat = analysis.beat
+        beat_intensity = analysis.beat_intensity
+
+        # ── Energy envelope (asymmetric smoothing) ──
+        alpha = self._energy_attack if rms > self._smoothed_energy else self._energy_release
+        self._smoothed_energy += alpha * (rms - self._smoothed_energy)
+        energy = min(1.0, self._smoothed_energy)
+        energy_delta = energy - self._prev_energy
+        self._prev_energy = energy
+
+        # ── Frequency band energy ──
+        bass_raw = (
+            float(np.mean(spectrum[:_BASS_END])) * self._sensitivity
+            if len(spectrum) > _BASS_END
+            else 0.0
+        )
+        mid_raw = (
+            float(np.mean(spectrum[_BASS_END:_MID_END])) * self._sensitivity
+            if len(spectrum) > _MID_END
+            else 0.0
+        )
+        treble_raw = (
+            float(np.mean(spectrum[_MID_END:])) * self._sensitivity
+            if len(spectrum) > _MID_END
+            else 0.0
+        )
+
+        a = self._band_smoothing
+        self._bass += a * (min(1.0, bass_raw) - self._bass)
+        self._mid += a * (min(1.0, mid_raw) - self._mid)
+        self._treble += a * (min(1.0, treble_raw) - self._treble)
+
+        # ── BPM estimation ──
+        if beat:
+            self._beat_times.append(now)
+            self._last_beat_time = now
+            # Keep only recent beats
+            if len(self._beat_times) > _MAX_BEAT_HISTORY:
+                self._beat_times = self._beat_times[-_MAX_BEAT_HISTORY:]
+
+        self._bpm = self._estimate_bpm()
+
+        # ── Beat phase (0-1 position in current beat cycle) ──
+        beat_phase = self._compute_beat_phase(now)
+
+        # ── Drop detection ──
+        self._update_drop_state(now, energy, energy_delta)
+
+        return MusicFeatures(
+            bpm=self._bpm,
+            beat=beat,
+            beat_intensity=beat_intensity,
+            beat_phase=beat_phase,
+            energy=energy,
+            energy_delta=energy_delta,
+            bass_energy=self._bass,
+            mid_energy=self._mid,
+            treble_energy=self._treble,
+            drop_state=self._drop_state,
+            drop_intensity=self._drop_intensity,
+        )
+
+    def _estimate_bpm(self) -> float:
+        """Estimate BPM from recent beat timestamps using median IBI."""
+        if len(self._beat_times) < 3:
+            return self._bpm  # keep previous estimate
+
+        # Compute inter-beat intervals
+        times = self._beat_times
+        ibis = [times[i] - times[i - 1] for i in range(1, len(times))]
+        # Filter out unreasonable IBIs
+        ibis = [ibi for ibi in ibis if 60.0 / _BPM_MAX <= ibi <= 60.0 / _BPM_MIN]
+        if not ibis:
+            return self._bpm
+
+        median_ibi = sorted(ibis)[len(ibis) // 2]
+        raw_bpm = 60.0 / median_ibi
+
+        # Exponential smoothing to avoid jitter
+        if self._bpm > 0:
+            return self._bpm * 0.85 + raw_bpm * 0.15
+        return raw_bpm
+
+    def _compute_beat_phase(self, now: float) -> float:
+        """Compute 0-1 phase within the current beat cycle."""
+        if self._bpm <= 0 or self._last_beat_time <= 0:
+            return 0.0
+
+        beat_period = 60.0 / self._bpm
+        elapsed = now - self._last_beat_time
+        phase = (elapsed % beat_period) / beat_period
+        return min(1.0, max(0.0, phase))
+
+    def _update_drop_state(self, now: float, energy: float, energy_delta: float) -> None:
+        """State machine: idle → buildup → drop → recovery → idle."""
+        state_duration = now - self._drop_state_time
+
+        if self._drop_state == "idle":
+            # Enter buildup when energy is rising and above minimum
+            if energy > _MIN_ENERGY_FOR_BUILDUP and energy_delta > 0.001:
+                self._drop_state = "buildup"
+                self._drop_state_time = now
+                self._buildup_start_energy = energy
+                self._peak_energy = energy
+
+        elif self._drop_state == "buildup":
+            self._peak_energy = max(self._peak_energy, energy)
+            # Transition to drop on sudden energy decrease
+            if self._peak_energy > 0 and energy < self._peak_energy * (1 - _DROP_THRESHOLD):
+                self._drop_state = "drop"
+                self._drop_state_time = now
+                self._drop_intensity = min(
+                    1.0, (self._peak_energy - energy) / max(self._peak_energy, 0.01)
+                )
+            # Cancel buildup if energy falls below minimum or state too long
+            elif energy < _MIN_ENERGY_FOR_BUILDUP or state_duration > 8.0:
+                self._drop_state = "idle"
+                self._drop_state_time = now
+
+        elif self._drop_state == "drop":
+            if state_duration >= _DROP_RECOVERY_TIME:
+                self._drop_state = "recovery"
+                self._drop_state_time = now
+
+        elif self._drop_state == "recovery":
+            # Fade drop intensity during recovery
+            progress = min(1.0, state_duration / _RECOVERY_DURATION)
+            self._drop_intensity = self._drop_intensity * (1.0 - progress)
+            if state_duration >= _RECOVERY_DURATION:
+                self._drop_state = "idle"
+                self._drop_state_time = now
+                self._drop_intensity = 0.0

server/src/wled_controller/core/processing/audio_stream.py

@@ -3,12 +3,17 @@
 Implements AudioColorStripStream which produces LED color arrays driven by
 real-time audio analysis (spectrum, beat detection, RMS levels).
 
-Three visualization modes:
-  spectrum    —  FFT frequency bars mapped across LEDs with palette coloring
-  beat_pulse  —  full-strip flash on beat detection with exponential decay
-  vu_meter    —  volume level fills LEDs like a progress bar
+Seven visualization modes:
+  spectrum        —  FFT frequency bars mapped across LEDs with palette coloring
+  beat_pulse      —  full-strip flash on beat detection with exponential decay
+  vu_meter        —  volume level fills LEDs like a progress bar
+  pulse_on_beat   —  BPM-synced pulsing with smooth beat phase
+  energy_gradient —  bass/mid/treble mapped to gradient with energy modulation
+  spectrum_bands  —  three VU zones for bass/mid/treble
+  strobe_on_drop  —  strobe effect triggered by detected musical drops
 """
 
+import math
 import threading
 import time
 from typing import Optional
@@ -18,6 +23,7 @@ import numpy as np
 from wled_controller.core.audio.analysis import NUM_BANDS
 from wled_controller.core.audio.audio_capture import AudioCaptureManager
 from wled_controller.core.audio.filters.pipeline import build_pipeline_from_template_ids
+from wled_controller.core.audio.music_analyzer import MusicAnalyzer
 from wled_controller.core.processing.color_strip_stream import ColorStripStream
 from wled_controller.core.processing.effect_stream import _build_palette_lut
 from wled_controller.utils import get_logger
@@ -68,6 +74,13 @@ class AudioColorStripStream(ColorStripStream):
         self._prev_spectrum: Optional[np.ndarray] = None
         self._prev_rms = 0.0
 
+        # Music analyzer (for semantic modes: pulse_on_beat, energy_gradient, etc.)
+        self._music_analyzer: Optional[MusicAnalyzer] = None
+        self._music_features = None  # latest MusicFeatures snapshot
+
+        # Strobe state
+        self._strobe_phase = 0.0
+
         self._gradient_store = None  # injected by stream manager
         self._update_from_source(source)
 
@@ -106,6 +119,17 @@ class AudioColorStripStream(ColorStripStream):
         self._auto_size = not source.led_count
         self._led_count = source.led_count if source.led_count and source.led_count > 0 else 1
         self._mirror = bool(getattr(source, "mirror", False))
+        self._beat_decay = bfloat(getattr(source, "beat_decay", 0.15), 0.15)
+
+        # Lazily create/update MusicAnalyzer for semantic modes
+        _MUSIC_MODES = {"pulse_on_beat", "energy_gradient", "spectrum_bands", "strobe_on_drop"}
+        if self._visualization_mode in _MUSIC_MODES:
+            if self._music_analyzer is None:
+                self._music_analyzer = MusicAnalyzer(sensitivity=self._sensitivity)
+            else:
+                self._music_analyzer.sensitivity = self._sensitivity
+        else:
+            self._music_analyzer = None
 
         # Resolve audio device/template via audio_source_id
         audio_source_id = getattr(source, "audio_source_id", "")
@@ -287,6 +311,10 @@ class AudioColorStripStream(ColorStripStream):
             "spectrum": self._render_spectrum,
             "beat_pulse": self._render_beat_pulse,
             "vu_meter": self._render_vu_meter,
+            "pulse_on_beat": self._render_pulse_on_beat,
+            "energy_gradient": self._render_energy_gradient,
+            "spectrum_bands": self._render_spectrum_bands,
+            "strobe_on_drop": self._render_strobe_on_drop,
         }
 
         try:
@@ -331,6 +359,10 @@ class AudioColorStripStream(ColorStripStream):
                         if analysis is not None and self._filter_pipeline is not None:
                             analysis = self._filter_pipeline.process(analysis)
 
+                        # Feed MusicAnalyzer for semantic modes
+                        if analysis is not None and self._music_analyzer is not None:
+                            self._music_features = self._music_analyzer.update(analysis)
+
                         render_fn = renderers.get(self._visualization_mode, self._render_spectrum)
                         t_render = time.perf_counter()
                         render_fn(buf, n, analysis)
@@ -477,3 +509,158 @@ class AudioColorStripStream(ColorStripStream):
         buf[:, 0] = r
         buf[:, 1] = g
         buf[:, 2] = b
+
+    # ── Music-sync modes (use MusicAnalyzer) ──────────────────────
+
+    def _render_pulse_on_beat(self, buf: np.ndarray, n: int, analysis) -> None:
+        """BPM-synced pulsing with smooth beat phase and exponential decay."""
+        if analysis is None or self._music_features is None:
+            buf[:] = 0
+            return
+
+        mf = self._music_features
+        beat_decay = self.resolve("beat_decay", self._beat_decay)
+
+        # On beat: flash to full brightness
+        if mf.beat:
+            self._pulse_brightness = 1.0
+        else:
+            self._pulse_brightness = max(0.0, self._pulse_brightness - beat_decay)
+
+        # Smooth sine pulse synced to BPM between beats
+        phase_pulse = 0.5 + 0.5 * math.sin(2.0 * math.pi * mf.beat_phase - math.pi / 2)
+        # Blend: sharp beat flash dominates, phase pulse fills gaps
+        brightness = max(self._pulse_brightness, phase_pulse * 0.3 * mf.energy)
+
+        if brightness < 0.005:
+            buf[:] = 0
+            return
+
+        palette_idx = max(0, min(255, int(mf.beat_intensity * 255)))
+        base_color = self._palette_lut[palette_idx]
+        buf[:, 0] = int(base_color[0] * brightness)
+        buf[:, 1] = int(base_color[1] * brightness)
+        buf[:, 2] = int(base_color[2] * brightness)
+
+    def _render_energy_gradient(self, buf: np.ndarray, n: int, analysis) -> None:
+        """Maps bass/mid/treble energy to gradient with energy-modulated brightness."""
+        if analysis is None or self._music_features is None:
+            buf[:] = 0
+            return
+
+        mf = self._music_features
+        lut = self._palette_lut
+
+        # Scroll gradient position based on beat phase
+        scroll = mf.beat_phase * 0.3  # subtle scroll
+
+        # Per-LED: position along gradient shifted by bass/treble balance
+        positions = np.linspace(0.0, 1.0, n, dtype=np.float32)
+        # Bass pushes warm, treble pushes cool
+        bass_shift = mf.bass_energy * 0.2
+        treble_shift = mf.treble_energy * 0.2
+        shifted = positions + scroll + bass_shift - treble_shift
+        np.mod(shifted, 1.0, out=shifted)
+
+        # Map to palette
+        indices = (shifted * 255).astype(np.int32)
+        np.clip(indices, 0, 255, out=indices)
+        np.copyto(buf, lut[indices])
+
+        # Modulate brightness by energy
+        brightness = max(0.05, mf.energy)
+        f32 = buf.astype(np.float32)
+        f32 *= brightness
+        np.clip(f32, 0, 255, out=f32)
+        np.copyto(buf, f32, casting="unsafe")
+
+    def _render_spectrum_bands(self, buf: np.ndarray, n: int, analysis) -> None:
+        """Three VU zones for bass/mid/treble with optional mirror."""
+        if analysis is None or self._music_features is None:
+            buf[:] = 0
+            return
+
+        mf = self._music_features
+        lut = self._palette_lut
+
+        if self._mirror:
+            # Mirror: bass in center, treble on edges
+            half = (n + 1) // 2
+            third = max(1, half // 3)
+            self._fill_band_zone(buf, 0, third, mf.treble_energy, lut, 200)
+            self._fill_band_zone(buf, third, 2 * third, mf.mid_energy, lut, 128)
+            self._fill_band_zone(buf, 2 * third, half, mf.bass_energy, lut, 40)
+            # Mirror second half
+            buf[half:] = buf[half - 1 :: -1][: n - half]
+        else:
+            third = max(1, n // 3)
+            self._fill_band_zone(buf, 0, third, mf.bass_energy, lut, 40)
+            self._fill_band_zone(buf, third, 2 * third, mf.mid_energy, lut, 128)
+            self._fill_band_zone(buf, 2 * third, n, mf.treble_energy, lut, 200)
+
+    @staticmethod
+    def _fill_band_zone(
+        buf: np.ndarray,
+        start: int,
+        end: int,
+        energy: float,
+        lut: np.ndarray,
+        palette_center: int,
+    ) -> None:
+        """Fill a zone of LEDs proportionally to energy level."""
+        zone_len = end - start
+        if zone_len <= 0:
+            return
+        fill = int(energy * zone_len)
+        buf[start:end] = 0
+        if fill > 0:
+            color = lut[max(0, min(255, palette_center))]
+            # Brightness gradient within the filled portion
+            brightness = np.linspace(0.4, 1.0, fill, dtype=np.float32)
+            for ch in range(3):
+                buf[start : start + fill, ch] = (color[ch] * brightness).astype(np.uint8)
+
+    def _render_strobe_on_drop(self, buf: np.ndarray, n: int, analysis) -> None:
+        """Strobe effect triggered by detected musical drops."""
+        if analysis is None or self._music_features is None:
+            buf[:] = 0
+            return
+
+        mf = self._music_features
+        color_main = self.resolve_color("color", self._color)
+        color_strobe = self.resolve_color("color_peak", self._color_peak)
+
+        if mf.drop_state == "idle":
+            # Gentle energy-modulated breathing
+            brightness = 0.1 + 0.3 * mf.energy
+            buf[:, 0] = int(color_main[0] * brightness)
+            buf[:, 1] = int(color_main[1] * brightness)
+            buf[:, 2] = int(color_main[2] * brightness)
+
+        elif mf.drop_state == "buildup":
+            # Increasing brightness pulsing
+            pulse = 0.5 + 0.5 * math.sin(time.perf_counter() * 8.0 * math.pi)
+            brightness = 0.3 + 0.7 * mf.energy * pulse
+            buf[:, 0] = int(color_main[0] * brightness)
+            buf[:, 1] = int(color_main[1] * brightness)
+            buf[:, 2] = int(color_main[2] * brightness)
+
+        elif mf.drop_state == "drop":
+            # Rapid strobe at ~10 Hz (capped for photosensitivity safety)
+            self._strobe_phase += 1
+            strobe_on = (self._strobe_phase % 3) < 2  # ~10 Hz at 30 fps
+            if strobe_on:
+                intensity = mf.drop_intensity
+                buf[:, 0] = int(color_strobe[0] * intensity)
+                buf[:, 1] = int(color_strobe[1] * intensity)
+                buf[:, 2] = int(color_strobe[2] * intensity)
+            else:
+                buf[:] = 0
+
+        elif mf.drop_state == "recovery":
+            # Fade from strobe back to breathing
+            fade = max(0.0, 1.0 - mf.drop_intensity)
+            brightness = 0.1 + 0.3 * mf.energy * fade
+            buf[:, 0] = int(color_main[0] * brightness)
+            buf[:, 1] = int(color_main[1] * brightness)
+            buf[:, 2] = int(color_main[2] * brightness)

server/src/wled_controller/core/processing/value_stream.py

@@ -186,9 +186,9 @@ class AudioValueStream(ValueStream):
         self._smoothing = smoothing
         self._min = min_value
         self._max = max_value
+        # auto_gain is deprecated — use the auto_gain audio filter instead.
+        # Field kept for backward compat but no longer has any effect.
         self._auto_gain = auto_gain
-        self._rolling_peak = 0.0  # tracks observed max raw audio value
-        self._rolling_decay = 0.995  # slow decay (~5-10s adaptation)
         self._audio_capture_manager = audio_capture_manager
         self._audio_source_store = audio_source_store
         self._audio_template_store = audio_template_store
@@ -288,12 +288,6 @@ class AudioValueStream(ValueStream):
 
         raw = self._extract_raw(analysis)
 
-        # Auto-gain: normalize raw against rolling observed peak
-        if self._auto_gain:
-            self._rolling_peak = max(raw, self._rolling_peak * self._rolling_decay)
-            if self._rolling_peak > 0.001:
-                raw = raw / self._rolling_peak
-
         raw = min(1.0, raw * self._sensitivity)
 
         # Temporal smoothing
@@ -340,18 +334,12 @@ class AudioValueStream(ValueStream):
             return
 
         old_source_id = self._audio_source_id
-        old_auto_gain = self._auto_gain
         self._audio_source_id = source.audio_source_id
         self._mode = source.mode
         self._sensitivity = source.sensitivity
         self._smoothing = source.smoothing
         self._min = source.min_value
         self._max = source.max_value
-        self._auto_gain = source.auto_gain
-
-        # Reset rolling peak when auto-gain is toggled on
-        if self._auto_gain and not old_auto_gain:
-            self._rolling_peak = 0.0
 
         # If audio source changed, re-resolve and swap capture stream
         if source.audio_source_id != old_source_id:

server/src/wled_controller/static/js/features/color-strips.ts

@@ -68,6 +68,7 @@ class CSSEditorModal extends Modal {
         if (_smoothingWidget) { _smoothingWidget.destroy(); _smoothingWidget = null; }
         if (_audioSensitivityWidget) { _audioSensitivityWidget.destroy(); _audioSensitivityWidget = null; }
         if (_audioSmoothingWidget) { _audioSmoothingWidget.destroy(); _audioSmoothingWidget = null; }
+        if (_audioBeatDecayWidget) { _audioBeatDecayWidget.destroy(); _audioBeatDecayWidget = null; }
         if (_effectIntensityWidget) { _effectIntensityWidget.destroy(); _effectIntensityWidget = null; }
         if (_effectScaleWidget) { _effectScaleWidget.destroy(); _effectScaleWidget = null; }
         if (_apiInputTimeoutWidget) { _apiInputTimeoutWidget.destroy(); _apiInputTimeoutWidget = null; }
@@ -122,6 +123,7 @@ class CSSEditorModal extends Modal {
             audio_source: (document.getElementById('css-editor-audio-source') as HTMLInputElement).value,
             audio_sensitivity: _audioSensitivityWidget ? JSON.stringify(_audioSensitivityWidget.getValue()) : '1.0',
             audio_smoothing: _audioSmoothingWidget ? JSON.stringify(_audioSmoothingWidget.getValue()) : '0.3',
+            audio_beat_decay: _audioBeatDecayWidget ? JSON.stringify(_audioBeatDecayWidget.getValue()) : '0.15',
             audio_palette: (document.getElementById('css-editor-audio-palette') as HTMLInputElement).value,
             audio_color: _audioColorWidget ? JSON.stringify(_audioColorWidget.getValue()) : '[]',
             audio_color_peak: _audioColorPeakWidget ? JSON.stringify(_audioColorPeakWidget.getValue()) : '[]',
@@ -166,6 +168,7 @@ let _cssTagsInput: any = null;
 let _smoothingWidget: BindableScalarWidget | null = null;
 let _audioSensitivityWidget: BindableScalarWidget | null = null;
 let _audioSmoothingWidget: BindableScalarWidget | null = null;
+let _audioBeatDecayWidget: BindableScalarWidget | null = null;
 let _effectIntensityWidget: BindableScalarWidget | null = null;
 let _effectScaleWidget: BindableScalarWidget | null = null;
 let _apiInputTimeoutWidget: BindableScalarWidget | null = null;
@@ -881,6 +884,19 @@ function _ensureAudioSmoothingWidget(): BindableScalarWidget {
     return _audioSmoothingWidget;
 }
 
+function _ensureAudioBeatDecayWidget(): BindableScalarWidget {
+    if (!_audioBeatDecayWidget) {
+        _audioBeatDecayWidget = new BindableScalarWidget({
+            container: document.getElementById('css-editor-audio-beat-decay-container')!,
+            min: 0.01, max: 0.5, step: 0.01, default: 0.15,
+            idPrefix: 'css-editor-audio-beat-decay',
+            valueSources: () => _cachedValueSources,
+            format: (v) => v.toFixed(2),
+        });
+    }
+    return _audioBeatDecayWidget;
+}
+
 function _ensureEffectIntensityWidget(): BindableScalarWidget {
     if (!_effectIntensityWidget) {
         _effectIntensityWidget = new BindableScalarWidget({
@@ -1267,12 +1283,16 @@ function _ensureAudioVizIconSelect() {
     const sel = document.getElementById('css-editor-audio-viz') as HTMLSelectElement | null;
     if (!sel) return;
     const items = [
-        { value: 'spectrum',   icon: _icon(P.activity),   label: t('color_strip.audio.viz.spectrum'),   desc: t('color_strip.audio.viz.spectrum.desc') },
-        { value: 'beat_pulse', icon: _icon(P.zap),        label: t('color_strip.audio.viz.beat_pulse'), desc: t('color_strip.audio.viz.beat_pulse.desc') },
-        { value: 'vu_meter',   icon: _icon(P.trendingUp), label: t('color_strip.audio.viz.vu_meter'),   desc: t('color_strip.audio.viz.vu_meter.desc') },
+        { value: 'spectrum',        icon: _icon(P.activity),   label: t('color_strip.audio.viz.spectrum'),        desc: t('color_strip.audio.viz.spectrum.desc') },
+        { value: 'beat_pulse',      icon: _icon(P.zap),        label: t('color_strip.audio.viz.beat_pulse'),      desc: t('color_strip.audio.viz.beat_pulse.desc') },
+        { value: 'vu_meter',        icon: _icon(P.trendingUp), label: t('color_strip.audio.viz.vu_meter'),        desc: t('color_strip.audio.viz.vu_meter.desc') },
+        { value: 'pulse_on_beat',   icon: _icon(P.heart),      label: t('color_strip.audio.viz.pulse_on_beat'),   desc: t('color_strip.audio.viz.pulse_on_beat.desc') },
+        { value: 'energy_gradient', icon: _icon(P.flame),      label: t('color_strip.audio.viz.energy_gradient'), desc: t('color_strip.audio.viz.energy_gradient.desc') },
+        { value: 'spectrum_bands',  icon: _icon(P.radio),      label: t('color_strip.audio.viz.spectrum_bands'),  desc: t('color_strip.audio.viz.spectrum_bands.desc') },
+        { value: 'strobe_on_drop',  icon: _icon(P.sparkles),   label: t('color_strip.audio.viz.strobe_on_drop'), desc: t('color_strip.audio.viz.strobe_on_drop.desc') },
     ];
     if (_audioVizIconSelect) { _audioVizIconSelect.updateItems(items); return; }
-    _audioVizIconSelect = new IconSelect({ target: sel, items, columns: 3 });
+    _audioVizIconSelect = new IconSelect({ target: sel, items, columns: 4 });
 }
 
 function _buildGradientEntityItems() {
@@ -1660,14 +1680,18 @@ function _resetMappedState() {
 
 export function onAudioVizChange() {
     const viz = (document.getElementById('css-editor-audio-viz') as HTMLInputElement).value;
-    // Palette: spectrum / beat_pulse
+    const MUSIC_MODES = new Set(['pulse_on_beat', 'energy_gradient', 'spectrum_bands', 'strobe_on_drop']);
+    // Palette: spectrum / beat_pulse / energy_gradient / spectrum_bands
     (document.getElementById('css-editor-audio-palette-group') as HTMLElement).style.display =
-        (viz === 'spectrum' || viz === 'beat_pulse') ? '' : 'none';
+        (viz === 'spectrum' || viz === 'beat_pulse' || viz === 'energy_gradient' || viz === 'spectrum_bands') ? '' : 'none';
     // Base color + Peak color: vu_meter only
     (document.getElementById('css-editor-audio-color-group') as HTMLElement).style.display = viz === 'vu_meter' ? '' : 'none';
     (document.getElementById('css-editor-audio-color-peak-group') as HTMLElement).style.display = viz === 'vu_meter' ? '' : 'none';
     // Mirror: spectrum only
     (document.getElementById('css-editor-audio-mirror-group') as HTMLElement).style.display = viz === 'spectrum' ? '' : 'none';
+    // Beat decay: new music modes only
+    (document.getElementById('css-editor-audio-beat-decay-group') as HTMLElement).style.display = MUSIC_MODES.has(viz) ? '' : 'none';
+    if (MUSIC_MODES.has(viz)) _ensureAudioBeatDecayWidget();
     _autoGenerateCSSName();
 }
 
@@ -1709,6 +1733,7 @@ function _loadAudioState(css: any) {
 
     _ensureAudioSensitivityWidget().setValue(css.sensitivity ?? 1.0);
     _ensureAudioSmoothingWidget().setValue(css.smoothing ?? 0.3);
+    _ensureAudioBeatDecayWidget().setValue(css.beat_decay ?? 0.15);
 
     const audioGradientId = css.gradient_id || 'gr_builtin_rainbow';
     (document.getElementById('css-editor-audio-palette') as HTMLInputElement).value = audioGradientId;
@@ -1729,6 +1754,7 @@ function _resetAudioState() {
     if (_audioVizIconSelect) _audioVizIconSelect.setValue('spectrum');
     _ensureAudioSensitivityWidget().setValue(1.0);
     _ensureAudioSmoothingWidget().setValue(0.3);
+    _ensureAudioBeatDecayWidget().setValue(0.15);
     (document.getElementById('css-editor-audio-palette') as HTMLInputElement).value = 'gr_builtin_rainbow';
     if (_audioPaletteEntitySelect) _audioPaletteEntitySelect.setValue('gr_builtin_rainbow');
     _ensureAudioColorWidget().setValue([0, 255, 0]);
@@ -1819,7 +1845,7 @@ const CSS_CARD_RENDERERS: Record<string, CardPropsRenderer> = {
     audio: (source, { audioSourceMap }) => {
         const vizLabel = t('color_strip.audio.viz.' + (source.visualization_mode || 'spectrum')) || source.visualization_mode || 'spectrum';
         const vizMode = source.visualization_mode || 'spectrum';
-        const showPalette = (vizMode === 'spectrum' || vizMode === 'beat_pulse') && source.palette;
+        const showPalette = (vizMode === 'spectrum' || vizMode === 'beat_pulse' || vizMode === 'energy_gradient' || vizMode === 'spectrum_bands') && source.palette;
         const audioPaletteLabel = showPalette ? (t('color_strip.palette.' + source.palette) || source.palette) : '';
         return `
             <span class="stream-card-prop">${ICON_MUSIC} ${escapeHtml(vizLabel)}</span>
@@ -2187,6 +2213,7 @@ const _typeHandlers: Record<string, { load: (...args: any[]) => any; reset: (...
                 audio_source_id: (document.getElementById('css-editor-audio-source') as HTMLInputElement).value || null,
                 sensitivity: _ensureAudioSensitivityWidget().getValue(),
                 smoothing: _ensureAudioSmoothingWidget().getValue(),
+                beat_decay: _ensureAudioBeatDecayWidget().getValue(),
                 gradient_id: (document.getElementById('css-editor-audio-palette') as HTMLInputElement).value,
                 color: _ensureAudioColorWidget().getValue(),
                 color_peak: _ensureAudioColorPeakWidget().getValue(),

server/src/wled_controller/static/locales/en.json

@@ -562,6 +562,7 @@
   "filters.inverter.desc": "Invert all levels (1 minus value)",
   "filters.beat_gate.desc": "Pass signal only around detected beats",
   "filters.delay.desc": "Time-shift the audio analysis by a delay",
+  "filters.auto_gain.desc": "Auto-normalize audio levels to use full range",
   "postprocessing.description_label": "Description (optional):",
   "postprocessing.description_placeholder": "Describe this template...",
   "postprocessing.created": "Template created successfully",
@@ -1264,6 +1265,16 @@
   "color_strip.audio.viz.beat_pulse.desc": "All LEDs pulse on the beat",
   "color_strip.audio.viz.vu_meter": "VU Meter",
   "color_strip.audio.viz.vu_meter.desc": "Volume level fills the strip",
+  "color_strip.audio.viz.pulse_on_beat": "Pulse on Beat",
+  "color_strip.audio.viz.pulse_on_beat.desc": "LEDs pulse with each detected beat",
+  "color_strip.audio.viz.energy_gradient": "Energy Gradient",
+  "color_strip.audio.viz.energy_gradient.desc": "Gradient intensity follows audio energy",
+  "color_strip.audio.viz.spectrum_bands": "Spectrum Bands",
+  "color_strip.audio.viz.spectrum_bands.desc": "Grouped frequency bands across the strip",
+  "color_strip.audio.viz.strobe_on_drop": "Strobe on Drop",
+  "color_strip.audio.viz.strobe_on_drop.desc": "Flash strobe effect on bass drops",
+  "color_strip.audio.beat_decay": "Beat Decay:",
+  "color_strip.audio.beat_decay.hint": "How quickly the beat pulse fades. Lower values = longer fade, higher = snappier response.",
   "color_strip.audio.source": "Audio Source:",
   "color_strip.audio.source.hint": "Audio source for this visualization. Can be a multichannel (device) or mono (single channel) source. Create and manage audio sources in the Sources tab.",
   "color_strip.audio.sensitivity": "Sensitivity:",

server/src/wled_controller/static/locales/ru.json

@@ -1194,6 +1194,16 @@
   "color_strip.audio.viz.beat_pulse.desc": "Все LED пульсируют в такт",
   "color_strip.audio.viz.vu_meter": "VU-метр",
   "color_strip.audio.viz.vu_meter.desc": "Уровень громкости заполняет ленту",
+  "color_strip.audio.viz.pulse_on_beat": "Пульс на бит",
+  "color_strip.audio.viz.pulse_on_beat.desc": "LED пульсируют при каждом ударе",
+  "color_strip.audio.viz.energy_gradient": "Энергетический градиент",
+  "color_strip.audio.viz.energy_gradient.desc": "Интенсивность градиента следует за энергией звука",
+  "color_strip.audio.viz.spectrum_bands": "Полосы спектра",
+  "color_strip.audio.viz.spectrum_bands.desc": "Группы частот по ленте",
+  "color_strip.audio.viz.strobe_on_drop": "Стробоскоп на дропе",
+  "color_strip.audio.viz.strobe_on_drop.desc": "Вспышка стробоскопа на басовых дропах",
+  "color_strip.audio.beat_decay": "Затухание бита:",
+  "color_strip.audio.beat_decay.hint": "Скорость затухания пульса. Меньшие значения = более долгое затухание, большие = более резкая реакция.",
   "color_strip.audio.source": "Аудиоисточник:",
   "color_strip.audio.source.hint": "Аудиоисточник для визуализации. Может быть многоканальным (устройство) или моно (один канал). Создавайте и управляйте аудиоисточниками на вкладке Источники.",
   "color_strip.audio.sensitivity": "Чувствительность:",

server/src/wled_controller/static/locales/zh.json

@@ -1194,6 +1194,16 @@
   "color_strip.audio.viz.beat_pulse.desc": "所有LED随节拍脉动",
   "color_strip.audio.viz.vu_meter": "VU 表",
   "color_strip.audio.viz.vu_meter.desc": "音量填充灯带",
+  "color_strip.audio.viz.pulse_on_beat": "节拍脉动",
+  "color_strip.audio.viz.pulse_on_beat.desc": "LED随每个检测到的节拍脉动",
+  "color_strip.audio.viz.energy_gradient": "能量渐变",
+  "color_strip.audio.viz.energy_gradient.desc": "渐变强度跟随音频能量",
+  "color_strip.audio.viz.spectrum_bands": "频谱频段",
+  "color_strip.audio.viz.spectrum_bands.desc": "分组频段分布在灯带上",
+  "color_strip.audio.viz.strobe_on_drop": "低音闪烁",
+  "color_strip.audio.viz.strobe_on_drop.desc": "低音下降时闪烁频闪效果",
+  "color_strip.audio.beat_decay": "节拍衰减：",
+  "color_strip.audio.beat_decay.hint": "节拍脉冲消退的速度。较低值 = 较长衰减，较高值 = 更灵敏的响应。",
   "color_strip.audio.source": "音频源：",
   "color_strip.audio.source.hint": "此可视化的音频源。可以是多声道（设备）或单声道（单通道）源。在源标签页中创建和管理音频源。",
   "color_strip.audio.sensitivity": "灵敏度：",

server/src/wled_controller/storage/color_strip_source.py

@@ -712,7 +712,8 @@ class AudioColorStripSource(ColorStripSource):
     LED count auto-sizes from the connected device when led_count == 0.
     """
 
-    visualization_mode: str = "spectrum"  # spectrum | beat_pulse | vu_meter
+    # spectrum | beat_pulse | vu_meter | pulse_on_beat | energy_gradient | spectrum_bands | strobe_on_drop
+    visualization_mode: str = "spectrum"
     audio_source_id: str = ""  # references an AudioSource (capture or processed)
     sensitivity: BindableFloat = field(default_factory=lambda: BindableFloat(1.0))
     smoothing: BindableFloat = field(default_factory=lambda: BindableFloat(0.3))
@@ -722,6 +723,7 @@ class AudioColorStripSource(ColorStripSource):
     color_peak: BindableColor = field(default_factory=lambda: BindableColor([255, 0, 0]))
     led_count: int = 0  # 0 = use device LED count
     mirror: bool = False  # mirror spectrum from center outward
+    beat_decay: BindableFloat = field(default_factory=lambda: BindableFloat(0.15))
 
     def to_dict(self) -> dict:
         d = super().to_dict()
@@ -735,6 +737,7 @@ class AudioColorStripSource(ColorStripSource):
         d["color_peak"] = self.color_peak.to_dict()
         d["led_count"] = self.led_count
         d["mirror"] = self.mirror
+        d["beat_decay"] = self.beat_decay.to_dict()
         return d
 
     @classmethod
@@ -753,6 +756,7 @@ class AudioColorStripSource(ColorStripSource):
             color_peak=BindableColor.from_raw(data.get("color_peak"), default=[255, 0, 0]),
             led_count=data.get("led_count") or 0,
             mirror=bool(data.get("mirror", False)),
+            beat_decay=BindableFloat.from_raw(data.get("beat_decay"), default=0.15),
         )
 
     @classmethod
@@ -777,6 +781,7 @@ class AudioColorStripSource(ColorStripSource):
         color_peak=None,
         led_count=0,
         mirror=False,
+        beat_decay=None,
         **_kwargs,
     ):
         return cls(
@@ -798,6 +803,7 @@ class AudioColorStripSource(ColorStripSource):
             color_peak=BindableColor.from_raw(color_peak, default=[255, 0, 0]),
             led_count=led_count,
             mirror=bool(mirror),
+            beat_decay=BindableFloat.from_raw(beat_decay, default=0.15),
         )
 
     def apply_update(self, **kwargs) -> None:
@@ -824,6 +830,8 @@ class AudioColorStripSource(ColorStripSource):
             self.led_count = kwargs["led_count"]
         if kwargs.get("mirror") is not None:
             self.mirror = bool(kwargs["mirror"])
+        if kwargs.get("beat_decay") is not None:
+            self.beat_decay = self.beat_decay.apply_update(kwargs["beat_decay"])
 
 
 @dataclass

server/src/wled_controller/templates/modals/css-editor.html

@@ -248,6 +248,10 @@
                             <option value="spectrum" data-i18n="color_strip.audio.viz.spectrum">Spectrum Analyzer</option>
                             <option value="beat_pulse" data-i18n="color_strip.audio.viz.beat_pulse">Beat Pulse</option>
                             <option value="vu_meter" data-i18n="color_strip.audio.viz.vu_meter">VU Meter</option>
+                            <option value="pulse_on_beat" data-i18n="color_strip.audio.viz.pulse_on_beat">Pulse on Beat</option>
+                            <option value="energy_gradient" data-i18n="color_strip.audio.viz.energy_gradient">Energy Gradient</option>
+                            <option value="spectrum_bands" data-i18n="color_strip.audio.viz.spectrum_bands">Spectrum Bands</option>
+                            <option value="strobe_on_drop" data-i18n="color_strip.audio.viz.strobe_on_drop">Strobe on Drop</option>
                         </select>
                     </div>
 
@@ -320,6 +324,17 @@
                         <div id="css-editor-audio-color-peak-container"></div>
                     </div>
 
+                    <div id="css-editor-audio-beat-decay-group" class="form-group" style="display:none">
+                        <div class="label-row">
+                            <label>
+                                <span data-i18n="color_strip.audio.beat_decay">Beat Decay:</span>
+                            </label>
+                            <button type="button" class="hint-toggle" onclick="toggleHint(this)" title="?" data-i18n-aria-label="aria.hint">?</button>
+                        </div>
+                        <small class="input-hint" style="display:none" data-i18n="color_strip.audio.beat_decay.hint">How quickly the beat pulse fades. Lower values = longer fade, higher = snappier response.</small>
+                        <div id="css-editor-audio-beat-decay-container"></div>
+                    </div>
+
                     <div id="css-editor-audio-mirror-group" class="form-group" style="display:none">
                         <div class="label-row">
                             <label for="css-editor-audio-mirror" data-i18n="color_strip.audio.mirror">Mirror:</label>

server/src/wled_controller/templates/modals/value-source-editor.html

@@ -134,17 +134,8 @@
                         </select>
                     </div>
 
-                    <div class="form-group">
-                        <div class="label-row">
-                            <label for="value-source-auto-gain" data-i18n="value_source.auto_gain">Auto Gain:</label>
-                            <button type="button" class="hint-toggle" onclick="toggleHint(this)" title="?" data-i18n-aria-label="aria.hint">?</button>
-                        </div>
-                        <small class="input-hint" style="display:none" data-i18n="value_source.auto_gain.hint">Automatically normalize audio levels so output uses the full range, regardless of input volume</small>
-                        <label class="toggle-label">
-                            <input type="checkbox" id="value-source-auto-gain">
-                            <span data-i18n="value_source.auto_gain.enable">Enable auto-gain</span>
-                        </label>
-                    </div>
+                    <!-- Auto-gain removed: use the auto_gain audio filter instead -->
+                    <input type="hidden" id="value-source-auto-gain">
 
                     <div class="form-group">
                         <div class="label-row">