Re-Performance Studio is the free-choice half of AVH. Where the
Re-Production Studio walks you through the whole song in one guided 8-step pass,
this page lets you jump straight to whichever part you want to re-perform — and run each one on its own.
Pick a module below: re-perform the instrumental with real instruments, re-perform the vocal
through your own trained voice model, or record a live sing-along of your real voice over the AI lead
(the Nuclear Option). Every path finishes with the same delivery-ready −14 LUFS / −1 dBTP master.
♪Re-Perform the AI CompositionRP-3 | 96 kHz | 24-bitTAKE 01PRO
1 Upload
2 Source
3 Perform
4 Master
5 Export
Tools
Path 1 — Instrumental re-performance
Humanize the bed in place, then de-fingerprint it. Timbre, melody and pitch are preserved — no re-synthesis, so the song never drifts off-key.
Starter + Pro
TranscribeAudio → MIDI note data (pitch, timing, velocity)Auto
HumanizeIn place — timbre & pitch preservedLocked
De-fingerprintSpectral ripple · 1–8 kHz band1.5
Fidelity gateCompared against your originalOn
MasterIntegrated loudness · true peak−14 LUFS
Take your AI composition and re-perform it with real instruments and human feel.
How it works▾
We transcribe your AI instrumental into MIDI, humanize the performance (timing, velocity, groove,
micro pitch drift), then re-render it through real-instrument models — so the finished track carries
the same musical content but a brand-new waveform with none of the AI generator's spectral fingerprint.
This applies to instrument stems only; vocals are handled by the main app's humanization chain.
1TranscribeAudio → MIDI note data (pitch, timing, velocity).
3Re-renderPlayed through real-instrument models, not the AI audio.
4MatchSpectral envelope, loudness & stereo width matched to the original.
1 Upload your audio
🎵
Drop an instrumental here or browse
WAV, MP3, FLAC, M4A · up to 200 MB
Leave this unchecked if you're uploading an instrumental stem — we’ll use it as-is for the cleanest re-performance. Tick it if you're uploading a full mix and only want the instrumental re-performed.
Recommended. The vocal is isolated and kept bit-for-bit, so the re-performed track still sounds like the same song. Untick to re-perform everything (including the vocal) as instruments.
2 Choose how to re-perform
Only the stems you tick are re-performed. The rest of the instrumental is kept exactly as-is and recombined with the re-performed parts.
3 Human feel & fidelity
🔒 LOCKED — in-key settings.
The instrumental is humanized in place (timbre & pitch preserved) and then de-fingerprinted.
These are the user-approved settings that keep the song in its original key, so the result never
drifts off-key. Re-synthesis (which goes off-key) is disabled.
4 Engine & fidelity upgrades auto-detected
These optional backends push the re-performance even closer to the original.
Anything unavailable in this deployment is greyed out and safely skipped.
Time-varying spectral match0.00
Tracks the original's brightness/loudness frame-by-frame instead of one average curve. Higher = closer to the original's dynamics.
5 Master the finished track
Recommended. Produces the same delivery-ready master as the Re-Production Studio's one-pass pipeline (ITU-R BS.1770 integrated loudness + a 4× oversampled true-peak limiter). Untick to keep the raw re-performed level.
Every re-performance is compared against your original to confirm the same music and the same instruments came back — not an off-key or mismatched cover.
Fidelity
—
⚠️ Flagged for review.
Re-Performance is a Pro feature. Upgrade to Pro ($24.99/mo) to take your AI composition and
re-perform it with real instruments and human feel — AUTO or CUSTOM, unlimited.
Path 2 — Vocal re-performance
Transcribe the AI lead, humanize its timing and phrasing, then re-render it through your own trained voice model — same words, melody and performance, brand-new human waveform.
Pro
TranscribeMelody, timing & phrasing of the AI leadAuto
HumanizeTiming jitter, breath & pitch micro-motion2.0 ms
Re-renderYour own voice model — self-hosted RVCOwn voice
Match originalEQ & space of the source vocalOn
Fidelity gateCompared against the original vocalOn
MasterIntegrated loudness · true peak−14 LUFS
Re-perform the singing through your own trained voice model.
How it works▾
We transcribe the AI vocal's melody, humanize its timing, phrasing and pitch micro-motion, then
re-render it through your own voice model — so the finished take carries the same words, melody and
performance, but a brand-new human waveform with none of the AI generator's vocal fingerprint. This is the
content-level re-performance that defeats the style sub-vector no signal-processing trick can touch.
3Re-renderThrough your own voice model (self-hosted RVC).
4MatchLevel-matched back into the arrangement.
1 Upload the vocal
🎤
Drop the vocal (or a full song) here or browse
WAV, MP3, FLAC, M4A · up to 200 MB
Tick this if you're uploading a full mix; we split out the lead vocal, re-perform it, then recombine it with the original bed. Leave it unchecked if you're uploading a vocal stem — we’ll use it as-is for the cleanest result.
2 Choose your voice model
Loading your voices…
Don't have a voice yet? Train one on the Voice & Likeness page — record a
30–60 second sample and we'll build a private model of your own voice. It's yours alone; nothing is
shared, and there's no per-use cost.
3 Human feel & voice conversion
4 Engine auto-detected
The self-hosted voice worker is used when available; otherwise a built-in preview engine runs.
5 Master the finished mix
Recommended. Applies the same delivery-ready master as the Re-Production Studio (ITU-R BS.1770 integrated loudness + a 4× oversampled true-peak limiter) to the recombined mix.
6 Match the original — tone & space recommended
Your re-performed vocal comes back dry and with the voice model's own tone. We'll seat it in the original AI vocal's EQ and space so it drops into the mix like the source — before you ever reach for the Vocal Enhancer or Atlas. Conservative and relative: if the original is dry, no reverb is added.
7 Vocal Polish — Enhancer & Atlas Reverb PRO
Your re-performed lead vocal comes back dry and close — the RVC re-render adds no space of its own. Shape it here with the same Vocal Enhancer suite used across the studios, including the new Atlas algorithmic reverb. When enabled, your polished vocal is what gets summed into the arrangement and mastered.
Your re-performed vocal is compared against the original to confirm the same words, melody and timing came back — not a mismatched or off-key take.
Fidelity
—
⚠️ Flagged for review.
Vocal Re-Performance is a Pro feature. Upgrade to Pro ($24.99/mo) to re-perform an AI vocal through
your own trained voice model — unlimited.
Path 3 — Live sing-along
Record your real voice over the AI lead. We time-align and pitch-match your take, low-pass it, and layer it under the lead so the two voices fuse into one human performance.
Pro
AlignYour take time-aligned & pitch-matched to the leadAuto
LayerTucked under the lead · 5 kHz low-pass−15 dB
De-fingerprintSpectral ripple · 1–8 kHz band1.5
Fidelity gateCompared against the original mixOn
MasterIntegrated loudness · true peak−14 LUFS
Sing it yourself. Record your real voice over the AI lead — the strongest de-fingerprint there is.
How it works▾
This is the Nuclear Option. We split your song into lead vocal, backing vocals and instrumental, then
time-align and pitch-match your live recording to the AI lead, low-pass it, and layer it a few dB under
the lead so the two voices fuse into one human take. We humanize the AI backing vocals, sum the whole vocal
bus, de-fingerprint it, and recombine it with the instrumental — then master the finished track to
−14 LUFS / −1 dBTP. The result keeps the AI song's words, melody and arrangement but carries your real
voiceprint, so no detector can match the generator's vocal fingerprint.
1SplitLead vocal, backing vocals & instrumental.
2AlignYour take time-aligned & pitch-matched to the lead.
3LayerBlended a few dB under the lead (the Nuclear Option).
4MasterFull mix mastered to −14 LUFS / −1 dBTP.
1 Bring your song
Already separated this song? Upload the lead vocal and instrumental and we’ll use them
exactly as they are — no re-listening and no separation artifacts, so the re-performance keeps
your original mix. The instrumental doubles as the arrangement waveform, so you don’t need the full song.
Choose a lead vocal and an instrumental to continue.
or drop the full song
🎵
Drop the full AI song here or browse
WAV, MP3, FLAC, M4A · up to 200 MB
We split the full song into lead vocal, backing vocals and instrumental automatically.
2 Break it into verse & chorus sections
Drag across the waveform to draw each verse and chorus, then drag the edges to line them up with the lead vocal lane. Double-click a section to loop it and practise your part before you record — exactly like the Collaboration Studio.
🎵 ArrangementNo sections yet
Fit
Song
Lead vocal
Backing vocals
Instrumental
Loading the waveform…
00:00 / 00:00
Drag across the waveform to draw a section. Drag its edges to resize, drag the middle to move it, and double-click a section to loop it and sing along. The Lead vocal lane shows exactly where the vocal starts and ends — line your verses and choruses up with it.
3 Cast the parts & pass the mic
Pick how you want to record, then pass the mic. Every take runs through the Vocal Enhancer below.
🎧System default
Line input — browser processing (AGC / noise suppression) is off.
🎧System default
🎥 One take — sing the whole song
Press Record and sing along to the lead vocal from the very start to the end of the song in one continuous take. It plays the song as your guide, runs your voice through the Vocal Enhancer, and when you stop it becomes your Real Vocal automatically.
Ready — press Record and sing along from the top.
TAKE 01one-take
0:00.0/0:00.0
00:00
…or upload a pre-recorded take instead
4 Record each verse & chorus
Record your real voice one section at a time — exactly like the Collaboration Studio. Save your arrangement above, then hit 🎛 Mixer on any verse or chorus to get a count-in and record just that part. Nailing one short section at a time is far easier than singing the whole song in a single pass. When you're done, assemble the takes into one continuous vocal and it flows straight into the render below.
🎬 Verse-by-verse takesSave your arrangement to begin
Draw your verses & choruses in Step 2, click Save arrangement, and they'll appear here ready to record.
5 Blend your take under the AI lead
Record a take above, or upload one, and this fader comes alive.
−15.0 dB
The classic tuck. Your voice adds air and human detail without ever taking the lead.
−40−24−15−90
Above −9 dB you are no longer tucking. Your room tone, mic self-noise and breath
now sit as loud as the lead's detail. That does not make the take harder to spot — it can
make it easier, because a second noise floor next to a clean one is a signature in itself.
If it sounds better up here, that is a real choice; just know what it costs.
—
WINDOW · WHOLE SONGLEAD + TAKE
LEAD
TAKE
AI lead—
Your take—
Sum—
GAIN −15.0 dB
LP 5.0 kHz
LAG —
NOT NORMALIZED
The preview is time-aligned only — it is not pitch-matched. Your take is shifted onto the
beat here and again when the parts are stitched, but pitch matching happens once, at the full-song
render. The preview is not loudness-normalized either. The −14 LUFS master happens at
render, so the balance you hear here is the balance you keep — but the overall level will come
up. Judge the relationship between the two voices, not the volume.
6 Processing
The Sing-Along path leaves your instrumental exactly as it is — nothing is re-performed, re-synthesised or re-timed, so the bed keeps its original performance and mix. Only the vocal is re-performed: your real voice replaces the AI lead.
Recommended. Produces the same delivery-ready master as the Re-Production Studio's one-pass pipeline.
7 Vocal Enhancer & Atlas Reverb PRO
Tune your live vocal in real time as you play the song — dynamics, tone, space, SFX, the signature Enhancer knob, and the new Atlas algorithmic reverb to place your lead vocal in a real space. It sits at the very end of the vocal chain, right before the final master.
Your finished song is compared against the original to confirm the same tempo, key and arrangement came back — with no dropouts from the re-performance.
Fidelity
—
⚠️ Flagged for review.
Live Sing-Along is a Pro feature. Upgrade to Pro ($24.99/mo) to record your real voice over the AI lead
and render the full mastered song — unlimited.
4count-in
Play it yourself. Plug a guitar, bass, keys or drums into your audio interface and play along.
How it works▾
Capture your instrument through a line-in on your audio interface (not the laptop mic) for a clean, low-noise
take. Play along with the track in your headphones while we record, then choose what happens next:
keep your live take as the instrument track, or blend it under the re-performed bed so your playing
sits underneath the music. We time-align your take to the song automatically, so a small round-trip latency
(browser + interface) is corrected for you.
1ConnectInstrument into a line-in on your interface.
2PlayMonitor on headphones and play along.
3AlignYour take time-aligned to the song.
4Keep / BlendUse it as-is, or sit it under the bed.
1 Upload the track to play along with
🎵
Drop the song or instrumental here or browse
WAV, MP3, FLAC, M4A · up to 200 MB
This is the backing track you'll hear in your headphones while you play. It also becomes the bed your take is aligned to.
2 Connect your instrument and play
Use headphones so the track doesn't bleed into your instrument take. Pick your instrument, choose the line-in device, check the input meter, then hit Record.
Your live take is compared against the backing track to confirm it lines up in time, key and chords — so it sits correctly with the song.
Fidelity
—
⚠️ Flagged for review.
Live Instruments are a Pro feature. Upgrade to Pro ($24.99/mo) to capture guitar, bass, keys or drums
through your interface and keep or blend your take — unlimited.
Compare two instrumentals and find out whether they’re the same song — and the same recording.
How it works▾
Upload two instrumentals. We run a forensic audio comparison — waveform correlation, chroma (harmony),
MFCC (timbre), spectrogram texture and the Chromaprint acoustic fingerprint (the algorithm behind AcoustID)
— and produce a full report with side-by-side waveform, spectrogram and chord images plus a
plain-language verdict on composition similarity, recording identity and copyright / content-ID risk.
1UploadTwo instrumentals (A and B).
2AnalyzeWaveform, harmony, timbre & fingerprint.
3CompareSame composition? Same recording?
4ReportFull forensic report + PDF.
1 Upload two instrumentals
A Instrumental A
Click or drop a file here
B Instrumental B
Click or drop a file here
If one of them is a full song with vocals, the analysis still works — the report isolates the instrumental bed automatically.
2 Run the comparison
Starting…
Instrumental Likeness Match is a Pro feature. Upgrade to Pro ($24.99/mo) to compare two instrumentals
and get the full forensic report.
5 Export the delivery master
Deliverables
Nothing to export yet. Run the re-performance on the Perform step and the
finished files will appear here.
Delivery spec
FormatWAV · PCM 16-bit
Sample rate44.1 kHz
Loudness−14 LUFS
True peak−1 dBTP
Every file is the same delivery-ready master the Re-Production Studio produces.
Create your account
Free to sign up. Choose a plan afterward to unlock Re-Performance.
Re-Performance Studio