WHERE WE'RE
HEADED
Keyboard clicks were just the beginning. From browser extensions across every platform to custom-trained AI models — we're building the universal noise elimination suite.
EQ Engine
Multi-band parametric EQ targeting 1-6 kHz keyboard click frequencies. Zero download, near-zero latency. The original Ripple Wave.
RNNoise ML Engine
Mozilla's RNNoise neural network bundled inside the extension. Broader noise suppression beyond keyboard clicks - fan hum, background hiss.
DeepFilterNet3 AI Engine
State-of-the-art deep learning model (~2MB, cached locally). Full-band filtering with audiophile-grade quality.
Auto-Detect & Channel Rules
Keyword + audio transient detection for clicky videos. Per-channel rules: always on, ask, or exclude. Subreddit rules on Reddit.
Reddit Support
Full support for Reddit video posts. Shadow DOM traversal for shreddit-player, SPA navigation via History API, subreddit-level rules.
Mouse Click Suppression
Targeted filtering for mouse click transients - different frequency profile from keyboard clicks but equally distracting in gaming/tutorial videos.
Per-Video Intensity Memory
Remember your intensity setting per channel or subreddit. Set HEAVY for a loud typer and LIGHT for a quiet keyboard - it sticks.
Firefox Extension
Port to Firefox using WebExtensions manifest. AudioWorklet and Web Audio API are fully supported. Same three engines, native Firefox addon store.
Safari Extension
Safari Web Extension using the Xcode template. Requires adapting the manifest and testing AudioWorklet support on WebKit. Full macOS + iOS Safari.
Twitch & Kick Support
Extend to Twitch.tv and Kick.com live streams. Hook into their HTML5 video players for real-time noise suppression during live gaming/coding streams.
Twitter/X, LinkedIn & Facebook Video
Support inline video players on X (Twitter), LinkedIn, and Facebook. Each platform uses a different video component - we handle them all.
Vimeo, Dailymotion & Bilibili
Expand to Vimeo, Dailymotion, and Bilibili (China's largest video platform). Standard HTML5 video hooking with platform-specific DOM selectors.
Universal Mode (Any HTML5 Video)
An "any page" toggle that hooks into any HTML5 video element on any website. Works on Coursera, Udemy, educational platforms, news sites, and more.
Fan & AC Noise Filter
Dedicated model for steady-state background noise - CPU fans, air conditioning, room hum. Spectral subtraction approach for minimal voice artifacts.
Adjustable Noise Profiles
Save and share custom filter presets. Build a profile for a specific creator whose setup has a distinct noise signature.
Multilingual Noise Models
Language-aware models optimized for tonal languages (Mandarin, Thai, Vietnamese), Indic languages (Hindi, Tamil), Arabic, Japanese, and Korean. Preserves language-specific pitch patterns that generic models damage.
RTL & CJK UI Localization
Full interface translation: Arabic, Hebrew (RTL layout), Chinese, Japanese, Korean, Hindi, Spanish, Portuguese, French, German, Russian, and more.
Custom Model Training Pipeline
Open-source training pipeline to fine-tune noise suppression models on your own data. Bring a dataset of noisy/clean pairs, get a model that runs natively in the extension.
RippleNet v1: Purpose-Built Model
Our own neural network trained specifically for keyboard/mouse click removal, not general-purpose denoising. Smaller, faster, and more accurate than repurposed models like RNNoise.
Adaptive Noise Fingerprinting
AI learns a creator's specific noise fingerprint from the first 5 seconds, then generates a session-specific suppression profile. Zero config, maximum quality.
Background Music Separation
Separate and optionally attenuate background music from the speaker's voice - great for cooking/vlog content where music competes with speech.
Echo & Reverb Removal
De-reverb to handle bathroom/bedroom recording setups. Some creators record in untreated spaces - this would make them sound like they're in a studio.
On-Device WASM Model Marketplace
Community marketplace for noise models. Creators and researchers publish models, users browse and install with one click. All models run on-device via WASM.
GOT A NOISE
TO SILENCE?
Vote for a feature, open an issue, or submit a PR. This is open source — your voice shapes the roadmap.