ReddRadar
Agent skill
wavecap-filters
Tune WaveCap audio filters and AGC settings. Use when the user wants to adjust highpass/lowpass filters, automatic gain control, de-emphasis, or improve audio quality for transcription.
Install this agent skill to your Project
npx add-skill https://github.com/majiayu000/claude-skill-registry/tree/main/skills/data/wavecap-filters
SKILL.md
WaveCap Audio Filter Tuning Skill
Use this skill to tune audio pre-processing filters that improve speech intelligibility before transcription.
Configuration Location
Audio filter settings are in the whisper: section of the config file:
- User config:
/Users/thw/Projects/WaveCap/state/config.yaml - Default config:
/Users/thw/Projects/WaveCap/backend/default-config.yaml
Filter Parameters
Highpass Filter (removes low-frequency rumble)
whisper:
highpassCutoffHz: 250 # Default: 250 Hz
| Value | Effect |
|---|---|
| 80-150 Hz | Preserves bass, some rumble may remain |
| 200-300 Hz | Balanced for emergency radio (default) |
| 400-500 Hz | Aggressive rumble removal, may thin voice |
When to adjust:
- Increase if you hear wind noise, engine rumble, or low-frequency hum
- Decrease if voices sound thin or tinny
Lowpass Filter (removes high-frequency hiss)
whisper:
lowpassCutoffHz: 3800 # Default: 3800 Hz
| Value | Effect |
|---|---|
| 3000-3500 Hz | More aggressive hiss removal, slightly muffled |
| 3800-4000 Hz | Balanced for radio speech intelligibility (default) |
| 4500-6000 Hz | Preserves more clarity, but keeps more hiss |
When to adjust:
- Decrease if there's persistent high-frequency hiss or static
- Increase if speech sounds muffled or lacks clarity
De-emphasis (compensates for radio pre-emphasis)
whisper:
deemphasisTimeConstantMicros: 75 # Default: 75 microseconds
| Value | Effect |
|---|---|
| 50 μs | Less treble reduction (North American FM standard) |
| 75 μs | Moderate treble reduction (default, good for most radio) |
| 100-150 μs | More treble reduction (European FM standard) |
When to adjust:
- If audio sounds too bright/harsh, increase the value
- If audio sounds too dull, decrease the value
Automatic Gain Control (AGC)
whisper:
agcTargetRms: 0.04 # Default: 0.04 (4% of full scale)
| Value | Effect |
|---|---|
| null | AGC disabled (use raw audio levels) |
| 0.02-0.03 | Light normalization, preserves dynamics |
| 0.04-0.05 | Moderate normalization (default) |
| 0.06-0.08 | Aggressive boost for very quiet sources |
When to adjust:
- Increase if audio is too quiet and Whisper misses words
- Decrease if audio clips or sounds distorted
- Set to
nullif source audio is already well-normalized
View Current Settings
grep -A20 "whisper:" /Users/thw/Projects/WaveCap/state/config.yaml | grep -E "(highpass|lowpass|deemphasis|agc)"
Common Tuning Scenarios
Noisy Environment (wind, engines, background noise)
whisper:
highpassCutoffHz: 350 # More aggressive low-cut
lowpassCutoffHz: 3500 # Reduce hiss
agcTargetRms: 0.05 # Boost speech above noise
Clean Studio/Dispatch Audio
whisper:
highpassCutoffHz: 150 # Preserve natural bass
lowpassCutoffHz: 4500 # Keep clarity
agcTargetRms: 0.03 # Light normalization
Weak/Distant Radio Signal
whisper:
highpassCutoffHz: 300 # Remove noise floor
lowpassCutoffHz: 3800 # Standard
agcTargetRms: 0.07 # Aggressive boost
Broadcastify/Web Streams (already processed)
whisper:
highpassCutoffHz: 200 # Light filtering
lowpassCutoffHz: 4000 # Preserve quality
agcTargetRms: 0.04 # Standard
Apply Changes
After editing state/config.yaml, restart the service:
launchctl stop com.wavecap.server && sleep 2 && launchctl start com.wavecap.server
Analyze Current Audio Quality
Check RMS levels of recent recordings to inform AGC tuning:
cd /Users/thw/Projects/WaveCap/backend && source .venv/bin/activate && python3 << 'EOF'
import numpy as np
from pathlib import Path
import struct
def get_wav_rms(filepath):
with open(filepath, 'rb') as f:
f.read(44) # Skip header
data = f.read()
samples = np.frombuffer(data, dtype=np.int16).astype(np.float32) / 32768.0
return np.sqrt(np.mean(samples**2))
recordings = sorted(Path("/Users/thw/Projects/WaveCap/state/recordings").glob("*.wav"),
key=lambda x: x.stat().st_mtime, reverse=True)[:10]
print("Recent recordings RMS levels:")
for r in recordings:
rms = get_wav_rms(r)
print(f" {r.name}: {rms:.4f} ({rms*100:.1f}%)")
EOF
Tips
- Start with default values and make small adjustments (10-20% changes)
- Test with a variety of recordings before committing to changes
- AGC is the most impactful setting for quiet/loud sources
- Highpass filter is most important for outdoor/mobile radio
- Keep notes on what works for different stream sources
Recommended Agent Skills
Expand your agent's capabilities with these related and highly-rated skills.
majiayu000/claude-skill-registry
agent-ops-spec
Manage specification documents in .agent/specs/. Use when user provides requirements, acceptance criteria, or feature descriptions that need to be tracked and validated against implementation.
majiayu000/claude-skill-registry
agent-ops-state
Maintain .agent state files. Use at session start, after meaningful steps, and before concluding: read/update constitution/memory/focus/issues/baseline consistently.
majiayu000/claude-skill-registry
agent-ops-spec
Manage specification documents in .agent/specs/. Use when user provides requirements, acceptance criteria, or feature descriptions that need to be tracked and validated against implementation.
majiayu000/claude-skill-registry
agent-ops-testing
Test strategy, execution, and coverage analysis. Use when designing tests, running test suites, or analyzing test results beyond baseline checks.
majiayu000/claude-skill-registry
agent-ops-testing
Test strategy, execution, and coverage analysis. Use when designing tests, running test suites, or analyzing test results beyond baseline checks.
majiayu000/claude-skill-registry
agent-ops-state
Maintain .agent state files. Use at session start, after meaningful steps, and before concluding: read/update constitution/memory/focus/issues/baseline consistently.
Didn't find tool you were looking for?