VibeSync Studio

10X No track loaded
ENGINE IDLE

Drop an audio file to begin

WAV, MP3, FLAC, M4A, OGG — analyzed locally in your browser. Nothing is uploaded.

Track

Load a track above to activate every tool
1.0×
Amp
—
0:00 / 0:00
Listening: Original Process the track to compare
Vol100%
PanC

Live Engine

Real-time meters while playing IDLE
Level (RMS)
— dB
True Peak
— dB
Short-term Loudness
— LUFS
Phase Correlation
—
-10+1

Real-Time Studio Rack

Live FX while playing — heard instantly, no render GR 0.0 dB

Original (A) never changes. Tweak live, or let AI Diagnose propose exact slider values → Preview → Apply to render a separate Processed (B). Use Match loudness for a fair A/B (so better ≠ just louder).

AI Diagnosis

Live Timeline

Sections · drops · key shifts (heuristic) — synced to playback —

Analysis

Auto-measured · no processing applied

Spectral balance

Mix & Master Diagnostics

AI Brain & Quality Score

Learns from every track you process
—quality
Tracks learned
0
Avg improvement
— pts
Preferred target
—
Best score yet
—
Process a few tracks and I'll start adapting Smart Master to what works for you.
Aggressive but safe restoration toward best achievable quality

Two-Track Match & Fix

Load two similar tracks · one click compares + fixes the weaker
Track A
drop or click to load
Track B
drop or click to load
Uses the higher-quality track as the reference

Stem Lab

Real AI separation via your server · honest mid/side fallback otherwise OFFLINE

No server? These buttons still work in-browser using mid/side math (karaoke-style vocal reduction + instrumental) — labelled as approximate. For true vocals/drums/bass/other, run the included vibesync-server and connect it.

Vocal Tone Shaper

Best on a separated vocal stem · works on the loaded track too READY

Advanced Noise Removal

Spectral denoise · hum notch · de-clip READY
Medium
Removes hiss/room/hum and tames clipping. Before/after via A/B.

Cover & Arrange Studio

Key · tempo · harmonies — real pitch/time engine READY

For your own recordings or material you have the rights to. Covers still need the songwriter's mechanical licence. This won't clone a specific person's voice, and it can't transcribe & rebuild every instrument from a library (that needs a server + multi-GB samples). Cover Modes below re-tone the same performance, so the musical feel is preserved.

Cover Modes — same performance, new vibe · feel-safe
40%
or transform pitch / time
Key change
0
Tempo change
0%
Harmony / backing vocals
45%

AI Cover & Modern Instrument Studio

Keep the song · modernize the music READY

Keeps your original composition, melody, timing & vocal and adds modern synthesized layers locked to the detected key, chords & BPM. Layers are synth (real Web Audio), not sampled instruments — and it can't transcribe & rebuild the original instruments (that needs a server + sample library). All layers are OFF by default; nothing is auto-applied or overwritten.

Instrument layers — off by default · pick what to add
50%

One-Click Cover keeps the vocal & adds pad/keys/bass. One-Click Instrumental removes the centre vocal (rough — use Stem Lab for a clean split) and rebuilds with modern layers. Result lands on B · Processed — compare with A, then 👍 Keep or 👎 revert.

Style packs — each generates 2 versions to choose from

Mastering Target

Pick a target, then Apply
Keep loudness (never go quieter)
ON: only enhances, never drops your level. OFF: hits the exact streaming target (may lower a loud master).

AI Decision Panel

What was found · what changed
No decisions yet. Run Smart Master or apply a target to see every change explained here.

Export

Renders the current A/B selection

Settings — AI Backend

Secure Gemini proxy · key stays on your server NOT SET

When set, the AI Brain talks to your own /api/ai-chat endpoint (Gemini). The API key is read from GEMINI_API_KEY on the server and never appears in this page. Leave empty during preview on claude.ai to use the built-in assistant.