YT-Levelr автор APMicro

YT-Levelr

An extension that automatically equalizes the volume of YouTube videos, especially podcasts where every producer has their own idea of what the correct levels should be.

YT Levelr is available for both Firefox and Chromium-based browsers.

Features

• Automatic Loudness Normalization: Automatically adjusts audio levels between videos to create a consistent listening experience
• Smart Gain Management: Uses asymmetric gain limits (more aggressive cuts than boosts) to prevent sudden loud blasts
• Gradual Confidence Growth: Gain limits start narrow and widen linearly until full adjustment range is available at 30s, allowing the algorithm to make small initial corrections before committing to its final decision
• Drift Correction: Slow automatic correction every 3 minutes after lock to handle long-term level shifts
• Noise Floor Protection: Ignores silence periods to avoid skewing measurements
• Local Processing Only: All audio processing happens entirely within your browser - no data is collected or transmitted
• Cross Browser Support: uses Manifest V3 for compatibility with chrome/edge/comet and firefox

How It Works

YT Levelr uses the Web Audio API to intercept audio output from YouTube's video element and applies automatic gain correction:

1. Audio Interception: Creates a source node from YouTube's video element
2. RMS Measurement: Continuously measures Root Mean Square (RMS) amplitude of the audio
3. Gain Calculation: Compares measured level against your target level (default: -22 dBFS)
4. Smooth Transitions: Applies gain changes with asymmetric attack/release times to protect against sudden loud sounds
5. Locking: After 30 seconds, gain locks and only drift correction applies

Strategy Details

• Gradual Confidence Growth:
• Starts with narrow gain limits (±6dB/+3dB) at t=0 for cautious initial corrections
• Limits widen linearly over time as confidence in measurement grows
• At 30s, full range is available (±20dB/+15dB) and gain locks to its final decision
• Asymmetric Treatment: Cuts are permitted more aggressively than boosts because a sudden loud blast is worse than staying quiet for a few seconds
• Gain Transitions: Faster transitions for cuts (0.15s) than boosts (1.2s) to protect against sudden loud audio

Optimized Content Types

• ✅ Podcasts: Primary use case - spoken word content works excellently
• ✅ Interviews: Conversational audio normalizes well
• ✅ Audiobooks: Consistent narration levels handled correctly
• ✅ Lectures: Educational content with steady delivery

Known Limitations

• ⚠️ Music Videos: Extended quiet intros may not be handled correctly due to the confidence-based approach
• ⚠️ Dynamic Range Content: Music with extreme dynamic range may benefit from manual adjustment
• ⚠️ Very Quiet Audio: Content consistently below -48 dBFS may not trigger measurements

Privacy

YT Levelr is privacy-first:

• ✅ No data collection or transmission
• ✅ All processing happens locally in your browser
• ✅ Only stores enabled and targetDB preferences locally