CH-001  ·  AI singing voice generator  ·  ON AIR

Any voice.
Your performance.

Phrasing, breath, micro-timing — preserved. Controlla Voice is an AI singing voice generator. Train an AI clone of your own singing from ten minutes of audio, swap into any voice you have rights to, generate full choirs from typed lyrics in thirty-plus languages.

Open the booth How it works
Web-based 30+ languages Royalty-free Voice cloning
Signal chain  ·  Three stages

From dry take
to finished voice.

No installation, no DAW required. Controlla Voice runs in the browser and routes your take through three steps. Your performance — phrasing, breath, timing — stays intact end to end.

STAGE · 01CH-001 / INPUT
01

Pick a voice or
upload your own.

Browse the Community Styles library or upload ten minutes of clean isolated singing to train your own AI singing voice. One credit per month on Plus, three on Creator. The trained model stays in your account permanently.

STAGE · 02CH-002 / TAKE
02

Upload a take or
type lyrics.

Drop in a WAV/MP3 vocal of any take you want re-voiced, or type lyrics straight into the choir generator. Toggle Clean it up to strip background noise and reverb before processing — works on most rough demos in one pass.

STAGE · 03CH-003 / OUTPUT
03

Get the voice.
Keep your take.

The output is your performance — every breath, every micro-timing — sung in the target voice. Export royalty-free for Spotify, Apple Music, YouTube, sync, anywhere. Voices and creations are private until you publish.

Signature capability

Train an AI clone of your own singing voice.

Ten minutes of clean, isolated vocal — no doubles, no reverb, no harmonies — is enough to train a Controlla model that captures the way you sing. After training, your model becomes a voice you can swap into any take, in any of the supported languages, even ones you don't speak.

  • Locked to your account — your voice clone stays yours; nothing is shared unless you publish or grant access.
  • Multilingual — sing in 30+ languages including ones you can't speak. The phrasing comes from your input take.
  • Royalty-free outputs — release commercially on Spotify, Apple Music, YouTube, sync deals, anywhere you'd publish a normal recording.
  • Ethical training only — Controlla requires user consent for every voice. Celebrity voices are prohibited without rights.
Train my voice
CH-008 / VOICE TRAININGREC
SOURCEisolated-vocal_take.wav
DURATION00 : 11 : 24
SAMPLESclean / no doubles
FIDELITYhigh — 48 kHz
STATUS▶ training · 67%
Operators  ·  Who tunes in

For people who sing,
write, and produce.

Controlla Voice is a working tool, not a toy demo. Four kinds of people pick it up and stay.

CH-101
SW

Songwriters

Demo a song in any vocal style before booking studio time. Hear your chorus in three different voices in an afternoon.

CH-102
PR

Producers

Fill the harmony stack, generate a choir from typed lyrics, or comp a guide vocal in the target tone before tracking the real take.

CH-103
VC

Vocalists

Sing parts beyond your range or in languages you don't speak. The input take is yours; the target timbre handles the rest.

CH-104
CR

Content creators

Type lyrics, pick a style, ship a vocal hook for a reel, podcast, ad, or short. Royalty-free outputs from the same export.

Patch bay  ·  Feature signal chain

What's wired in.

Specimen-by-specimen list of what Controlla Voice does inside the browser. No plugin install, no DAW required.

CH-201 · Voice swap

Same singer.
Different voice.

Drop a vocal take in, pick a voice, get the same performance back with the target timbre. Phrasing, breath, micro-timing, and the human feel of the take are preserved — most AI tools flatten that out.

CH-202 · Choir gen

Text in.
Choir out.

Type lyrics, pick a style, get a layered choir in seconds. Use Community Styles or roll your own with My Styles.

CH-203 · Voice clone

10 min of audio,
your own model.

Train your own singing voice from a short clean recording. Locked to your account, permanent, multilingual.

CH-204 · Clean it up

Rough demo?
One toggle.

Built-in cleanup strips background, reverb, and noise from the input so the swap reads cleanly. Two-pass results rare.

CH-205 · Privacy

Private
by default.

Voices and outputs are visible only to you. Grant explicit access to collaborators, producers, or engineers per project.

CH-206 · 30+ languages

Sing in languages
you don't speak.

Generate a vocal in Japanese, Spanish, Korean, Yoruba, or twenty-plus others. The take you uploaded — phrasing, intent — carries through. The pronunciation comes from the target language model.

ENESFRDEJAKOZHPTITHIYO+ 20
Specimen comparison

Where Controlla
fits — and where
it doesn't.

Controlla Voice sits in the same room as Suno, Kits.ai, ElevenLabs and the open-source RVC family — but it does a narrower thing than most. Read this before subscribing.

Controlla Voice Suno Kits.ai ElevenLabs RVC (open)
Voice swap, performance preserved Yes No Yes Speech-first Yes (self-host)
Full song generation No Yes No No No
Train your own voice (no code) Yes — 10 min audio No Yes Speech only Code required
Text → choir generator Yes Via full song No No No
Browser-based, no install Yes Yes Yes Yes Self-host
30+ languages including unknown ones Yes Yes Limited 29 — speech Model dependent
Free tier or open source Subscription Free credits Free tier Free quota 100% free
Honest: if you want to type a prompt and get a finished song, Suno. If you want voice training for free and have the patience to self-host, RVC. Controlla is the right call when you have your own take and want it in a different voice without losing the performance.
Listener log  ·  Field reports

Three takes,
one mixed review.

Selected from producer forums and songwriter Discords. The 4-star one stays in on purpose.

★ ★ ★ ★ ★ CH-301

"Demoed a chorus in three voices before lunch. Booked the right singer in the afternoon."

M
Maya R.Songwriter · Nashville
★ ★ ★ ★ ☆ CH-302

"Voice swap is excellent. The monthly credit limit on custom training makes a long album a planning exercise."

J
Julian K.Producer · Berlin
★ ★ ★ ★ ★ CH-303

"Typed Japanese lyrics, picked a soprano style, exported a vocal hook in under a minute. Reel went live the same day."

A
Anna T.Content creator · Lisbon
Provenance

A music tech startup,
not a foundation model.

Controlla is a music-tech startup whose product suite includes Voice Clone, Voice Swap, and Collab. Voice — what this site is about — is the singing-voice generator. The company's public position on AI training has been consistent since launch: every voice model requires explicit consent from the source vocalist, and using a celebrity's likeness without rights is prohibited at the platform level, not just the policy level.

The thing Controlla actually does well — and what we'd point you to it for — is preserving the human feel of a take while changing the voice. Most AI voice tools flatten phrasing and breath into something that sounds like everyone. Controlla's swap keeps the input performance intact and only re-paints the timbre.

What we'd flag honestly: the custom voice-training feature uses a credit system — one trained model per month on Plus, three on Creator — so if you're cycling through experimental voices, plan ahead. The output is very sensitive to input quality; rough demos work, but a clean, dry, single-layer vocal will always swap better. There's no native DAW plugin yet, so the workflow is browser → export → import.

For the official product surface, plan details, and the training requirements page, see voice.controlla.xyz.

STATION LOGFM 88.4
CompanyControlla
Product suiteVoice · Swap · Collab
Voice releasedEarly access · ongoing
HostingWeb app
PlansBasic · Plus · Creator
Voice training min.10 min isolated audio
Languages30+
Royalty-freeYes — commercial OK
Privacy defaultPrivate to user
Celebrity voicesProhibited without rights
Caller queries  ·  Frequently asked

Things people
actually ask.

What exactly does Controlla Voice do?
Controlla Voice is an AI singing-voice generator. You hand it an input — either a recorded vocal take or typed lyrics — and pick a target voice. The output is a vocal sung in that target voice while keeping the rhythm, phrasing, breath, and timing of your input. It also trains custom AI voices from your own recordings, generates layered choirs from typed lyrics, and supports thirty-plus languages. It does not generate full instrumental songs the way Suno does — only voice.
Is everything I make royalty-free for commercial release?
Yes. Outputs you make through Controlla can be released commercially — Spotify, Apple Music, YouTube, sync, advertising, all of it — provided the input voice is one you have the rights to use (your own, or one with permission). The platform does not charge per-stream royalties on Controlla-generated output.
How much audio is needed to train my own voice?
Ten minutes of clean, isolated singing is the published minimum, and up to about an hour is supported. The recording should be a single-layer vocal: no doubles, no harmonies, no reverb, no instrumentals in the same file. Cleaner inputs produce stronger models. Controlla provides a guided recording flow for first-time users without studio access.
Can I generate a vocal in a language I don't speak?
Yes — that's one of the platform's headline features. The phrasing and intent come from the take you upload; the pronunciation and accent come from the target language model. Thirty-plus languages are supported, including languages you've never spoken. Pronunciation accuracy depends on the target language model's training depth.
Can I use Controlla Voice inside my DAW?
There is no native VST3/AU plugin today. Controlla runs in the browser. The standard workflow is: render or upload your take to Controlla, run the swap or choir generation, export the result, and bring the file back into your DAW. We'd love a plugin and so would the community — at the time of writing, this is the most common feature request in the Discord.
How does the pricing work?
Plans are Basic, Plus, and Creator. Basic does not support custom voice training. Plus includes one custom voice credit per month and unlocks commercial release. Creator gives three custom voice credits per month and the largest conversion quota. Voice credits are used when training a new model, not when running swaps on an existing trained voice.
Is my voice safe? Who else has access to my recordings and models?
Voices and outputs are private to your account by default. You can explicitly grant access to specific collaborators — producers, songwriters, engineers — per project, and revoke that access at any time. Controlla's published position is that nothing is shared or published without an explicit action from you. Your trained voice clone remains yours even if you cancel your subscription.
Can I clone a celebrity's voice or a famous singer?
No, and the platform actively prevents it. Controlla's terms of service require explicit consent from the source vocalist for every trained voice. Using a celebrity's likeness or a specific named artist's voice without rights is prohibited at the platform level and accounts found doing so are suspended. This is a deliberate position, not a workaround.
How does Controlla compare to Suno, ElevenLabs, or RVC?
Different jobs. Suno generates full songs from text prompts. ElevenLabs is excellent for speech but not specifically tuned for sung performance. RVC is a free open-source voice-conversion project that requires self-hosting and command-line comfort. Controlla sits in the singing-specific voice-swap niche with a strong stance on consent and a UI built for non-technical producers and songwriters. Kits.ai is the closest direct competitor; pick based on which voice library and pricing fits your project.
Booth ready  ·  ON AIR

Open the booth.
Sing in any voice.

Web-based, no install. Bring a take or type lyrics — pick a voice, get the part. Your performance stays yours. Royalty-free on export.

Open the booth Train my voice
Web-based · 30+ languages · Royalty-free · Voice cloning · Choir generator
Controlla Voice
FM 88.4 · On Air
Open