Get a professionally trained RVC or So-VITS voice model built from your own audio dataset — delivered ready to use on your PC for unlimited voice cloning.
I Will Train a Custom RVC or So-VITS Voice Model From Your Audio
A single trained RVC or So-VITS voice model from your audio dataset.
- Train one RVC or So-VITS model on your provided audio
- RVC output: .pth model file + .index file
- So-VITS output: config.json + .pth file + .pt file
- Minimum ~5 minutes of high-quality audio accepted
- Files delivered ready for local PC inference
- Unlimited local use of the trained model after delivery
Enhanced training with extended dataset support and a longer training run for improved voice quality.
- Everything in Starter, at a higher training scope
- Optimised for larger or more varied audio datasets
- Extended training duration for improved model accuracy
- RVC output: .pth model file + .index file
- So-VITS output: config.json + .pth file + .pt file
- Files delivered ready for local PC inference
Full-scope training run for the most detailed and natural-sounding voice model output.
- Everything in Boost, at maximum training scope
- Highest epoch count for the most refined voice reproduction
- Best suited for 20-minute-plus high-quality datasets
- RVC output: .pth model file + .index file
- So-VITS output: config.json + .pth file + .pt file
- Priority handling within the queue
Request a Custom Offer
Log In to Request a Custom Offer
Create a free account or log in to request a personalised offer from this Zinner.
Log In / RegisterAsk a Pre-Sale Question
Log In to Ask a Question
To reduce platform spam, pre-sale messages can only be sent by logged-in users.
Create a free account or log in to message this Zinner directly.
Log In / RegisterAt a Glance
Key details about this service to help you decide. Generated by Zinn Hub, not the seller.
Value Position
Model Format
Audio Requirement
Deliverable Ownership
Turnaround
What You'll Receive
Full Description
If you want to clone a voice with precision and reuse it endlessly on your own machine, you need a properly trained model — not a rushed, low-epoch attempt. This service delivers a fully trained RVC or So-VITS voice model built directly from the audio you provide, giving you complete ownership of the output and the freedom to run it locally whenever you need it.
Voice cloning technology has advanced rapidly, and RVC (Retrieval-based Voice Conversion) alongside So-VITS-SVC are two of the most capable open-source frameworks available today. Getting the training right — correct preprocessing, appropriate epoch counts, and properly indexed output files — makes the difference between a convincing clone and a muddy, robotic result. That is exactly what this service focuses on.
**What You Receive**
For RVC model training, you receive a `.pth` model file and a `.index` file — everything you need to load the voice into any compatible RVC interface on your PC.
For So-VITS model training, you receive a `config.json`, a `.pth` model file, and a `.pt` file — the complete set required to run inference locally.
Once delivered, you can use these files on your own computer to convert audio into the cloned voice as many times as you wish, with no ongoing fees or platform dependency.
**How the Process Works**
1. You supply your audio dataset — ideally 20 minutes or more of clean, high-quality recordings of the target voice. A minimum of around 5 minutes is workable if the audio quality is excellent and consistent.
2. Training begins on your dataset using the appropriate framework (RVC or So-VITS, whichever you select).
3. You receive the completed model files via the order, ready to load into your local setup.
**Who This Is For**
This service suits musicians, producers, content creators, developers, and hobbyists who want to work with a specific voice in their projects — whether that is their own voice, a character voice, or any voice for which they hold the necessary rights and permissions. It is also ideal for anyone exploring AI audio tools who wants a properly trained foundation model rather than a generic preset.
**Why This Seller**
The training is handled with care for dataset quality and output integrity. You are not simply handed a half-trained checkpoint — the process is run through properly, and you receive the full file set needed to use the model immediately. Clear communication is available throughout via order chat should any questions arise about your dataset or deliverables.
Zinner Quality Guarantee
Every Zinner is reviewed and approved before joining the platform.
All services are backed by our quality assurance commitment.
Your payment is protected until you approve the delivered work.
Compare Packages
| Feature | Starter | Boost | Premium |
|---|---|---|---|
| Delivery Time | 2 days | 3 days | 4 days |
| Revisions | 0 | 0 | 0 |
| Train one RVC or So-VITS model on your provided audio | ✓ | ✕ | ✕ |
| RVC output: .pth model file + .index file | ✓ | ✓ | ✓ |
| So-VITS output: config.json + .pth file + .pt file | ✓ | ✓ | ✓ |
| Minimum ~5 minutes of high-quality audio accepted | ✓ | ✕ | ✕ |
| Files delivered ready for local PC inference | ✓ | ✓ | ✕ |
| Unlimited local use of the trained model after delivery | ✓ | ✕ | ✕ |
| Everything in Starter, at a higher training scope | ✕ | ✓ | ✕ |
| Optimised for larger or more varied audio datasets | ✕ | ✓ | ✕ |
| Extended training duration for improved model accuracy | ✕ | ✓ | ✕ |
| Everything in Boost, at maximum training scope | ✕ | ✕ | ✓ |
| Highest epoch count for the most refined voice reproduction | ✕ | ✕ | ✓ |
| Best suited for 20-minute-plus high-quality datasets | ✕ | ✕ | ✓ |
| Priority handling within the queue | ✕ | ✕ | ✓ |
Portfolio
Examples of the seller's work related to this Zinn.
Train a Custom RVC or So-VITS Voice Model From Your Audio
Train a Custom RVC or So-VITS Voice Model From Your Audio
Extra Information
My Process
Perfect For
Tools I Use
Frequently Asked Questions
RVC (Retrieval-based Voice Conversion) is generally easier to get running locally and is widely supported by community tools. So-VITS-SVC can produce very natural results but requires a slightly more involved local setup. If you are unsure, mention your use case in the order chat and guidance will be provided.
A minimum of around 5 minutes of clean, consistent, high-quality audio is workable. However, 20 minutes or more will yield a noticeably more accurate and natural-sounding model. Background noise, music, or heavy reverb in the source audio will reduce output quality, so clean recordings are strongly preferred.
Once you place your order, upload your audio files directly via the order chat or use a cloud storage link shared in the chat. Accepted formats include WAV, FLAC, and MP3. WAV or FLAC at 44.1 kHz or higher is recommended for best results.
The delivered files are standard model checkpoints compatible with RVC and So-VITS community interfaces that run locally on Windows or Linux PCs with a CUDA-capable GPU. You will need to have the relevant inference software installed on your machine. A setup guide addon is available if you need assistance getting started.
Model training is a computational process rather than a creative one — the output quality is directly tied to the dataset you provide. If there is a technical issue with the delivered files (e.g. a corrupted or incomplete file), this will be resolved at no extra charge. If you wish to retrain with a different dataset or settings, please get in touch via order chat to discuss options.
You are responsible for ensuring you hold the appropriate rights and permissions to train a model on the audio you provide. This service is intended for use with your own voice, original recordings, or audio for which you have explicit authorisation.
The Starter tier is suitable for straightforward voice cloning with a clean dataset of around 5–20 minutes. The Boost tier is recommended when you have a larger or more varied dataset and want a more refined result. The Premium tier is ideal for the highest-quality output, particularly with 20 minutes or more of audio and when accuracy is critical.
Customer Reviews
See what our customers say about this Zinn
Thank you! It came out great:D ⭐️⭐️⭐️⭐️⭐️
Thanks again! Sounds awesome! :D
I was extremely satisfied with the delivery of my vocal model order. The process was quick and smooth, with clear communication at every step. The product arrived in perfect condition, exactly as described. I highly recommend this service for its professionalism and efficiency.
Only logged in customers who have purchased this product may leave a review.
