The simplest AI voice workspace
One-tap RVC training turns your audio into a voice model you own —
.pth, .index and ZIP. Convert vocals with it, then separate
stems, remove noise and remove reverb, all on one page.
- You own the files. Every voice you train downloads as .pth, .index and ZIP and works anywhere RVC runs - no platform can take it away
- No installation, notebook, or local GPU
- Explore paid options if you need more
Training is working normally again. Earlier today more people were training at once than the GPUs we had available could handle, and some training jobs failed for a few hours. Apologies if yours was one of them - please try it again.
Active trainings
Selected job
Training details
Epoch progress comes directly from the running RVC trainer.
An inference-ready early model is available while training continues.
It has been validated, backed up, and added to your model library.
Trained on your phone? Send the files somewhere you can use them.
The provider returned an error. Your other saved models were not affected.
Voice training
Train an RVC voice from an audio file
Already have an RVC model? Skip the training and convert a vocal with it. NiceVois conversions are designed to reduce robotic s, sh, f, t, and breath artifacts. Hear standard RVC versus NiceVois.
Voice conversion
Convert a vocal or full song with your voice
Convert a vocal or a full song with a saved NiceVois voice or your own RVC model. Every conversion includes the normal RVC output and a NiceVois Clean Consonants version for more natural sibilance and breaths.
Don’t have a voice yet?
Train your voice nowCheck this for a full song, or for a vocal with harmonies layered over it. We split the lead out, convert only the lead, and mix it back with the untouched music and harmonies. Leave it unchecked if your file is already a clean solo vocal.
Converts ten seconds of your file at every pitch from −12 to +12, so you can hear which one suits this voice.
Choose a voice and audio on the left, then press Convert.
Voice models buzz on the sounds that carry no pitch: the s, f, sh, ch and t sounds, plus the breaths between lines. We put the original vocals back in those spots, so the singing changes voice and the consonants stay human.
DownloadBefore you train
Prepare your audio
A voice model can only be as good as the recording behind it. These three clean up the usual problems — music under the voice, background noise, and room echo — and each one is free to try.
Stem separator
Split a song into vocals and instrumental
Start from a song, come away with the voice on its own and the instrumental on its own.
Previous runs
Noise remover
Take hiss, hum and room noise off a recording
Hiss, hum and room noise come off. What was said or sung stays exactly as it was.
Previous runs
Reverb remover
Take room and echo off a vocal you want dry
The room's wash comes off and the recording comes back dry, ready to place anywhere.
Previous runs
Trim
Cut the silence and the parts you don't want trained
Drag across the waveform, press Backspace or Delete, and the two sides join without a click. Or let it find every silent gap for you.
Vocal mixing
Add echo, reverb and space to a vocal
Studio echo and reverb, applied to the voice only. Nothing else in the track is touched.
Format converter
Convert an audio file to another format
WAV, MP3, FLAC, M4A, OGG, Opus, AIFF, WMA and more. The audio itself is left exactly as it is.
Pricing
Training balance and conversion plans.
One plan. Every tool.
Your first training is free - then pay only for the cloud training you use. Every account also includes five voice-conversion download minutes.
Your first training is free. After that, one subscription covers every tool on this page with nothing to meter - and every model you train is still a file you download and keep. Cancel any time; it stays yours.
- Up to 500 epochs for every balance-funded training.
- An exact quote first based on your audio and selected epochs.
Training balance
Pay only when you train
One balance for every NiceVois training — voice models and image models alike. Your balance never expires. For more voice-conversion downloads, choose a monthly plan below.
20% off your first balance purchase. Applied automatically at checkout.
Your first training
Create a free account and your first voice training is free. It also comes with five voice-conversion download minutes, and ten minutes each on the stem separator, noise remover and reverb remover.
Checking eligibility...$5 training balance
One-time, not a subscription. At current quotes, $5 covers up to 11 one-minute tests at 100 epochs, 4 five-minute runs at 180 epochs, or 2 ten-minute runs at 180 epochs. The exact cost and balance left are shown before training.
$10 training balance
Pay different amounts for short tests and longer, stronger training.
$27.50 training balance
Best for creators training several voices or comparing epoch settings.
Monthly conversion plans
More voice-conversion download minutes
Use the included free minutes first. Subscribe only when you need more conversion downloads; monthly plans also include renewable training credit.
Pricing
Pay once, or subscribe
Buy one training or three, with nothing to cancel. The monthly plan covers every tool if you use them often.
25% off your first month. Regular monthly price after that.
Train one voice
One training run, any length, up to the full epoch range. Five minutes of conversion downloads included. No subscription, nothing to cancel.
Train three voices
Three training runs, any length, up to the full epoch range. Fifteen minutes of conversion downloads included. No subscription, nothing to cancel.
Every tool, one price
Every tool, unlimited: voice training, voice conversion, stem separation and cleanup.
For occasional projects
$5 monthly training credit and 30 voice-conversion download minutes.
For regular cover makers
$10 monthly training credit and 90 voice-conversion download minutes.
For high-volume work
$20 monthly training credit and 240 voice-conversion download minutes.
Secure checkout by Lemon Squeezy Cancel any time from your account Your model files are yours to keep
Before you pay
- Do I need an account to pay?
- No. Pay with any email. Sign in with that same email on any device and everything you paid for is there.
- What does unlimited actually mean?
- No minutes to count on any tool, every output included, up to 500 epochs per training run. Fair use applies.
- How do I cancel?
- One click from your account, no email needed. You keep everything until the end of the month you paid for.
- What if a training fails?
- You are not charged for it. A failed or cancelled run releases what it reserved, and an upload that never arrives does not spend your free run.
- Who sees my card?
- Lemon Squeezy, the merchant of record. NiceVois never sees card details, and tax is handled at checkout.
- Do I keep my voices if I leave?
- Yes. Every model is a .pth, .index and ZIP you download, and they work anywhere RVC runs.
- What happens to my audio?
- It is processed for your job and never shared. Source audio is removed after validation; finished models stay until you delete them.
- Can I hear it before I pay?
- Yes. Your first training is free, and every finished model comes with a listening proof before you decide anything.
Wrong charge or an unused month? Email info@nicevois.com with the checkout email and date; it is reviewed with Lemon Squeezy and returned to the original payment method. Refund policy · Privacy
Secure checkout is being prepared.
Persistent cloud storage
Your model library
Your voice models are files you own. Download them once and no platform can take them away.
Your .pth, .index, ZIP, and private listening preview remain available after closing the browser or restarting your computer. Full-resolution source audio and temporary training files are removed after validation.
Checking private cloud storage…
Simple by design
Drop a file, press once, download the result
No notebook, commands, local setup, or GPU. Every tool on this page works the same way, and each one runs on its own - you never have to train a voice to use the others.
- 1Choose your tool
Train a voice, convert a vocal or full song, split stems, remove noise, or remove reverb.
- 2Drop your audio and press once
NiceVois runs it on a cloud GPU. You can close the page; training and conversion keep going.
- 3Download the files
Stems, cleaned audio, converted vocals, or a trained voice as .pth, .index and ZIP - yours to keep.
Your files, not ours
Everything you make here is yours to keep
NiceVois is not a hosted voice platform. A trained voice is a file - .pth, .index and a ZIP - that you download once and use anywhere RVC runs. Cancel, leave, or forget about us, and it still works.
- PortableStandard RVC v2 files. They load in Applio, RVC WebUI and any runtime that reads a .pth - not only here.
- Not tied to an accountThe subscription pays for the work. It is never the price of keeping your own files.
- Not at a platform's mercyWhen Weights.gg shut down, hosted models went with it. A file on your own disk cannot be shut down.
Explore NiceVois
Tools and guidance for AI vocal work
Go straight to the input you have, or learn how to prepare, train, and use the files you receive.
Start with the input you have
Why the operation is model training rather than a file-format rename.
After trainingRVC voice converter and vocal cleanerHear how NiceVois cleans robotic consonants, sibilance, and breaths.
Whole-song workflowChange the voice in a Suno songSee every step from a finished song and permitted voice to a downloadable cover.
MP3 inputMP3 to PTH for RVCWhen compressed audio is useful and what quality problems to check.
WAV inputWAV to PTH for RVCHow lossless source audio moves through the cloud training pipeline.
One-tap workflowTrain an RVC model onlineUpload, configure once, and let NiceVois handle the training workflow.
Guides See all guides ›
What happens between upload and a ready-to-use model package.
DatasetHow much audio do you need?Duration targets and what to cut before you upload.
Dataset qualityAudio requirements for RVC trainingFormats, duration, isolation, and the problems that weaken a model.
Training strengthHow many RVC epochs should you use?Use audio duration and urgency to choose a practical starting point.
Model filesRVC .pth vs .index explainedWhat each output contains and why the complete package includes both.
After trainingHow to use your RVC modelWhere the .pth and .index belong, and how to run your first conversion.
CoversMake an AI cover with your modelSeparate, convert, set the pitch, and mix it back.
Model formatsPTH to ONNX explainedWhen conversion helps, what it drops, and when to skip it.
New workflow experiment
Upload 8–40 images and keep the portable .safetensors package, without Colab or a local GPU.
Image datasetPrepare images for LoRA trainingWhich variations help, what to remove, and why duplicates weaken the result.
Coming from another service
Whether your old models are recoverable, and how to rebuild one.
Focused alternativeReplacing Weights.gg trainingWhat this utility replaces, and what it deliberately does not.
No local GPUAn Applio alternativeThe same RVC v2 output without an NVIDIA card or a Python setup.
Own the fileA Kits.ai alternativeWhen you need a portable model rather than a hosted voice library.
Custom voiceA Jammable alternativeTrain your own voice instead of picking one from a catalogue.
Real trainingA Musicfy alternativeA full RVC training and the model files, rather than a 30-second clone.
About NiceVois
Verified artifacts, pipeline checks, observed runtimes, and limitations.
Observed benchmarkTwo simultaneous 50-epoch runsSee the measured 4m34s and 5m42s end-to-end beta results.
Previous NiceVois serviceLooking for lyric-changing services?The original personalized-song service is still available through Fiverr.
Common questions
Common questions
What can NiceVois do?
Five things, all on this page: train a voice model from your audio, convert a vocal or a full song with a voice, split a song into vocals and instrumental, remove noise from a recording, and remove reverb from a vocal.
Do I own the model?
Yes. Every voice you train is delivered as a portable .pth and .index that you download and keep. They load in any RVC v2 runtime - Applio, RVC WebUI and others - and keep working whether or not you keep a NiceVois account or subscription. Nothing is hosted-only.
Do I have to train a voice to use the other tools?
No. The stem separator, noise remover and reverb remover take any audio file and need no voice model. Conversion needs a voice - either one you trained here or your own .pth you upload.
What does training look like?
Upload your audio, choose the epochs, name your model, and press Start Training. NiceVois handles everything else and returns a downloadable .pth, matching .index, and complete ZIP.
Is this literally converting audio into a .pth file?
It is training an RVC voice model from the audio. A .pth is the resulting model-weight file, so the operation requires dataset preparation and GPU training rather than a simple format conversion.
Which audio formats can I upload?
Every tool accepts WAV, MP3, FLAC, M4A, and OGG. For training, clean dry voice audio with minimal background sound produces the most reliable result.
How much audio should I use?
Two to fifteen minutes of clean, single-speaker audio is the practical range, and recordings shorter than one minute may not produce a reliable model. NiceVois shows an exact quote from the total audio length and epochs before training starts.
Do I need an NVIDIA GPU or local RVC installation?
No. Training runs on an external cloud GPU. Once training has started, you can close the page or shut down your computer and return later.
What can I download?
You can download the named .pth by itself, the .index by itself, or a ZIP package containing both. When available, an early inference-ready checkpoint can also be downloaded while the final run continues.
How long are beta files stored?
Public-beta model files and their private compressed listening previews are available for seven days. Full-resolution source audio and temporary training files are removed after validation. Download the package before its displayed expiry date.
Now available
Start free. Every tool.
Train a voice free, and try the converter, separator, noise remover and reverb remover before you decide.
One free training per person, up to 150 epochs and under 5 minutes of audio. Balance-funded runs add longer audio and up to 500 epochs.
One free training per person, up to 150 epochs and under 5 minutes of audio. A subscription adds longer audio and up to 500 epochs.
Add non-expiring training balance. Each run shows its workload price upfront.
One subscription covers training, conversion, separation and cleanup, with no limits and every download included.
Get NiceVois product updates.
Hear about major training improvements, new capabilities, and availability updates.
From NiceVois customers