The short answer
RVC, Retrieval-based Voice Conversion, is the underlying technology: a way to train a model on one voice and convert other vocals into it. The RVC WebUI is the original project's interface for training and converting. Applio is an independent, community-driven app, released under the MIT license, that builds on RVC and adds its own interface, desktop apps, and extra features.
Because they share the same foundation, an RVC model trained in one usually works in the other. The real differences are setup effort, polish, and what sits around the core conversion.
The RVC WebUI
The WebUI is the reference implementation. It runs as a local web page after you install Python and its dependencies, and it exposes everything: dataset processing, training, index building, conversion with every setting, model merging, and checkpoint extraction. Its wiki FAQ is still one of the best explanations of epochs, dataset size, and the index rate.
It is built for people comfortable with a terminal. Updates are less frequent than the community forks, and most guides assume Windows with an NVIDIA graphics card.
Applio
Applio, from the IAHispano community, packages RVC into something closer to a consumer app. It offers a desktop app for Windows, Mac, and Linux, plus the Python and Gradio version for people who want it. Its Mac builds cover both Apple Silicon and Intel on macOS 12 and later.
On top of conversion and training it adds text to speech, real-time conversion, and a model search. It is free and open source, and its community is active. The main caveat on Mac is that the app is not Apple-notarized yet, so first launch needs a one-line Terminal fix that Applio documents on its download page.
Side by side
Setup: the WebUI needs a manual Python install; Applio offers installable desktop apps. Platforms: the WebUI is happiest on Windows with NVIDIA; Applio ships Windows, Mac, and Linux builds. Features: both train and convert; Applio adds text to speech and real-time conversion; the WebUI keeps the original merge and checkpoint tools. Cost: both are free and open source.
Models: both use the same .pth and .index files, so you can switch without retraining. Pay attention to the RVC generation, v1 or v2, and the sample rate, which matter more than which app made the file.
What about the other forks?
RVC has spawned many community forks over the years, and Applio itself grew out of that scene, building on earlier forks such as Mangio-RVC-Fork. Plenty of older forks and one-click installers are no longer maintained, and they are where most broken installs come from today.
If a tutorial tells you to download a fork you have never heard of, check when its code was last updated and whether its issues page is active. An abandoned fork can still work, but you will be on your own when it breaks after a system update.
Which one should you use?
If you like to tinker and want the original behavior, use the WebUI. If you want RVC for free with less setup and an active community, Applio is an excellent choice, and we would recommend it to anyone happy with open-source tools and their occasional rough edges.
If you want RVC as one part of a music workflow rather than a standalone tool, that is where Xport Studio fits. It runs the same kind of RVC conversion and training locally, alongside stem splitting, noise removal, key and BPM detection, and mastering, with a Library that keeps every voice organized. It needs no Python, and the Mac build is notarized. Voice conversion is free for 7 days, then part of Xport Pro, $79 one time; the other tools are free. See voice cloning and the Voice Blender.
Moving models between them
Copy the .pth file and its matching .index file into the new tool. In the WebUI they go in the weights and logs folders; in Applio and Xport you load them through the interface. If a model behaves differently after moving, check the pitch extraction method and the index rate before blaming the file; our guide to the best RVC settings walks through them.
Want something new from models you already own? You can merge two RVC models into an original voice in either the WebUI or Xport's Voice Blender.