Sonic Lab · Live workspace
Conversational Audio & Video Studio
Record and monitor voice, apply live tone and dynamics, edit vocals and videos, and export your work. Browser processing needs no paid AI keys.
Ready. You control when your devices and files are accessed.
Live audio processor
Use headphones while monitoring to avoid feedback. Voice thickening is tonal processing, not identity cloning.
Download processed recordingLive video preview
Camera activation always requests browser permission. Filters apply to your local preview; video recording is supported where MediaRecorder is available.
Download videoUse the live tools above, or select a dedicated workstation. Audio and video recordings are not uploaded automatically.
Lead + Backing Vocal Swap Studio
Upload individually separated audio tracks. Keep or replace the lead, the backing singers, or both. Everything in this browser editor stays on your device.
1. Load tracks
Need separated stems? The included optional Demucs CLI separates music from vocals. Lead-versus-backing vocals must be supplied separately or generated with an appropriate specialized separator.
2. Route and mix
3. Conversational vocal control
Commands: “keep backing and replace lead”, “replace both”, “mute lead”, “swap backing from 12 to 25”, “lead gain -3”, “export wav”.
Ready.
4. Video soundtrack replacement
After mixing, choose “Render WebM” to replace the audio in your video using the generated mix. Supported browsers require canvas capture and MediaRecorder.
Use only recordings and voice identities you own or have permission to process. Neural voice identity conversion needs an independently configured model; this editor performs waveform routing and mixing.
Offline Live Media Engine
Work on files locally in your browser. No API keys or upload required. Large files depend on browser memory.
Audio editor
Video and image lab
WebM exports are silent unless you select a soundtrack below. Browser support for MediaRecorder and canvas streams is required.
Conversational media controller
Try “trim 3 to 15”, “gain 4”, “warm vocals”, “fade in 2”, “normalize on”, “master audio”, “video brightness 130”, “title Sunset”, “render video”. Commands change visible controls; processing happens on explicit export.
Vocal & Instrumental Quality Lab
Preserve the original, add weight to thin vocals, enhance instrumentals, and compare the result before exporting 24-bit WAV. Processing runs in your browser.
1 · Source
2 · Quality controls
3 · Conversational controls
Ready. No upload needed.
4 · Quality report
RMS is a basic level indicator, not integrated LUFS. A peak-safe render is not a certified true-peak master. The 24-bit WAV preserves the render resolution, not information missing from the original.
Source audio stays on your device. Keep the unprocessed original. Heavy processing can reduce natural dynamics or expose recording artifacts; use before/after listening.
Media modules
Note: browser-based processing is real and local; true neural text-to-video, universal media editing and AI voice identity conversion require separately installed compatible models, rights to source material, and server resources.