``Last update: March 14, 2026`` *** ###### ‎ :::content-center ## Introduction ‎ :icon-book: ::: - Mainline RVC is the base, **original**, & unmodified official version of RVC. Made by the [RVC-Project](https://github.com/RVC-Project) team. It can be called either Original/Mainline RVC. - It has less features compared to other [forks](https://docs.aihub.gg/essentials/whats-rvc/#forks), but still has the necessary tools to do a decent job. - It's specially liked because it's a little faster than other forks, as it's less bloated in a way. - Its actual name is not "Mainline", but it was given by the public to properly distinguish it from the other versions. ‎ #### Pros & Cons :icon-tasklist: ==- ***Unfold*** !!! *The pros & cons are subjective to your necessities.* !!! ||| ✔️ **PROS** - Easy to install. - Simpler to use. ||| ❌ **CONS** - Doesn't have an active development. - Has less features. - Manual model upload. - Won't work for RTX 50 Series GPUs. - Older dependencies. ||| === *** ###### ‎ :::content-center ## Installing & Opening Nvidia on Windows (Precompiled) :icon-download: ::: ###### ‎ 1. Go to their download page [here](https://huggingface.co/lj1995/VoiceConversionWebUI/blob/main/RVC1006Nvidia.7z) 3. Click the ``Download`` word. RVC will begin to download. ‎ image ‎ 3. Once it's done, unzip the folder. 4. Open RVC's folder, find the "``go-web.bat``" file and execute it. ‎ image ‎ It will then open a console, & after a moment your default web browser with RVC ready to be used. ‎ ‎ image‎ ‎ 5. **(Optional)** To access RVC more easily, make a shortcut of the ``go-web`` file. ‎ !!!warning Don't close the console until you are done using RVC, or it will stop working. !!! *** ###### ‎ :::content-center ## Inference :icon-unmute: !!!success If you encounter an issue, be sure to read the [Troubleshooting](https://docs.aihub.gg/rvc/local/mainline/#troubleshooting-) chapter. !!! ::: ###### ‎ #### 1. Upload voice model. a. Open RVC's folder, go to the `assets` folder and put your model's [**.PTH**](https://docs.aihub.gg/essentials/voice-models/#voice-model-files) file inside the `weights` folder. ‎ image ‎ b. Return to the previous folder & put the model's [**.INDEX**](https://docs.aihub.gg/essentials/voice-models/#voice-model-files) file in the `logs` folder. image *** ###### ‎ #### 2. Select voice model. a. In RVC, click the ``Refresh voice list and index path`` button. image ‎ b. In its left, click `Inferencing voice` & select your model. image *** ###### ‎ #### 3. Select vocals. In ``Enter the path of the audio file`` paste the [path file](https://static1.howtogeekimages.com/wordpress/wp-content/uploads/2023/09/shift-right-click-copy-as-path.png?q=50&fit=crop&w=767&dpr=1.5) of your audio. Ensure the path doesn't include spaces or special characters. image *** ###### ‎ #### 4. Modify settings. (optional) If you wish, modify the [inference settings](https://docs.aihub.gg/rvc/resources/inference-settings/) on display accordingly for better results. *** ###### ‎ #### 5. Convert. Click the long ``Convert`` button at the bottom & it will begin to convert. The processing time will mainly depend on your specs, length of audio, & the algorithm picked. *** ###### ‎ #### 6. Download output. Once it's done processing, a playable audio will pop up in the `Export audio` box. To download, click the three dots on the right & hit `Download`. image‎ *** ###### ‎ :::content-center ## Training :icon-dependabot: ###### ‎ !!!warning NOTES: The training guide will be centered around using [TensorBoard](https://docs.aihub.gg/rvc/resources/training/#tensorboard). Read about it first if you haven't already. If you encounter an issue, be sure to read the [Troubleshooting](https://docs.aihub.gg/rvc/local/mainline/#troubleshooting-) chapter. !!! ::: ###### ‎ :::content-center ### Step 1 ::: ###### ‎ #### 1. Go to training area. Open RVC & head over to the `Train` tab. image *** ###### ‎ #### 2. Name the model. In `Enter the experiment name` you insert a name for your model. Don't include special characters or spaces. image *** ###### ‎ #### 3. Select Target Sample Rate. In `Target sample rate` select the number that matches your datasets' [sample rate](https://docs.aihub.gg/rvc/resources/dataset-isolation/#step-1-find-the-sample-rate). Inputting an incorrect one might screw up the final quality. image *** ###### ‎ :::content-center ### Step 2a ::: ###### ‎ #### 4. Select dataset. In `Enter the path of the training folder` paste the [path file](https://static1.howtogeekimages.com/wordpress/wp-content/uploads/2023/09/shift-right-click-copy-as-path.png?q=50&fit=crop&w=767&dpr=1.5) of your dataset. Ensure the path doesn't include special characters/spaces. image ‎ !!!secondary If there's any text in the bar, delete it beforehand. !!! *** ###### ‎ #### 5. Process data. Click the `Process Data` button on the center. RVC will process the previous criteria for the training. But also the dataset file, which might take a moment depending on how big it is. image ‎ It'll finish when the output box on the right says ``end preprocess``. *** ###### ‎ :::content-center ### Step 2b ::: ###### ‎ #### 6. Select GPUs. In `Enter the GPU index(es)` determine which GPU(s) you'll use for training, by indicating the index followed by the dash (e.g: `0`). image *** ###### ‎ #### 7. Select pitch extraction algorithm. a. At the right select the [**Pitch extraction algorithm**](https://docs.aihub.gg/rvc/resources/inference-settings/#pitch-extraction-algorithm). Only use ``RMVPE_GPU`` or ``Crepe``, as the rest are obsolete. image ‎ ###### ‎ b. Now click the `Feature extraction` button on the right. image‎ ‎ ‎ It'll finish when the output says ``all-feature-done``. image *** ###### ‎ #### 8. Create .INDEX. Press `Train feature index` at the bottom center. This will create the [.INDEX](https://docs.aihub.gg/essentials/voice-models/#voice-model-files) file. image‎ ‎ It'll finish when the output box says something like this: image *** ###### ‎ :::content-center ### Step 3 ::: ###### ‎ #### 9. Select save frequency. Frequency of the [saving checkpoints](https://docs.aihub.gg/extra/glossary/#checkpoints), based on the [epochs](https://docs.aihub.gg/rvc/resources/training/#tensorboard). If you are a newbie, simply leave it at `15`. E.g: with a value of ``10``, they will be saved after the epoch 10, 20, 30, etc. ‎ image *** ###### ‎ #### 10. Input epochs amount. In `Total training epochs` you determine the total amount of [epochs](https://docs.aihub.gg/rvc/resources/training/#epochs--overtraining) (training cycles) for the model. But since we'll use [TensorBoard](https://docs.aihub.gg/rvc/resources/training/#tensorboard), use an arbitrarily large value like `2000`. image *** ###### ‎ #### 11. Select batch size. Leave `Batch size per GPU` at `8` if you aren't familiar with it. If your dataset is short (around 2 minutes or less), use ``4`` instead. image *** ###### ‎ #### 12. Launch TensorBoard. Now before you start training, open TB. If you haven't already, start reading about it here [here](https://docs.aihub.gg/rvc/resources/training/#tensorboard). *** ###### ‎ #### 13. Begin training. Start training the model by clicking `Train model`. image‎ ‎ ‎ Remember to monitor TB, & also the console just in case. The latter will show you errors if they happen, and information about the epochs & checkpoints. ‎ image *** ###### ‎ #### 14. Stop training. When you are very sure of overtraining, you can stop training by pressing the `Stop training` button where `Train model` used to be. *** ###### ‎ #### 15. Gather model's files. a. Create a new folder anywhere named as your model. a. Open RVC's folder, go to ``logs``, and open the folder named with the model. Select the `.INDEX` named ``added_`` & move it to your newly made folder. image ‎ c. Now go to the ``weights`` folder. Here you'll find the model's [checkpoints](https://docs.aihub.gg/extra/glossary/#checkpoints). Select the one **closest** to ***before*** the overtraining point, and move it to the new folder These files will be organized with this format: **ModelName_Epoch_Step.pth** Example: ``kalomaze_e60_s120.pth`` image ‎ ‎ And that's all. Have fun with your model. To test the model, do a normal [inference](https://docs.aihub.gg/essentials/how-to-make-ai-cover/) as usual. *** ###### ‎ :::content-center ### Resuming :icon-sync: ::: ###### ‎ If the training finished but the model still needed training, you don't have to start from scratch. **Follow this procedure:** - Simply enter the **same settings and criteria** that you previously inserted. Model name, sample rate, dataset, batch size, etc. You don't have to press ``Process Data`` or train the [.INDEX](https://docs.aihub.gg/essentials/voice-models/#voice-model-files) again. - You can change the **save frequency**, or increase the **epochs** amount in case you didn't input enough before. - Begin training again & remember to monitor TB & console like before. *** :::content-center ## Troubleshooting :icon-tools: ::: ==- :icon-alert: Important: HAGS on Windows 10/11 If you are experiencing lag, stuttering, or slow training speeds, **disable HAGS** (Hardware-Accelerated GPU Scheduling) on Windows 10/11. It is known to interfere with VRAM management for Local AI apps. [Read the HAGS Glossary Entry](https://docs.aihub.gg/extra/glossary/#hags-hardware-accelerated-gpu-scheduling) for the full explanation and how to disable it. === ==- There's no option for my sample rate. - **If it's lower than 32k**: select ``32k``. - **If it's 32k**: on the right in Version, press `v1` & press `v2` again. Ensure you leave it as `v2`. You should be able to see a `32k` option now. - **If it's 44.1k**: select ``40k``. - **If i'ts higher than 48k**: select ``48k``. === ==- The voice glitches out. - This a phenomenon called artifacting. To fix it, read [here](https://docs.aihub.gg/rvc/resources/dataset-isolation/#artifacts). === ==- I don't see the Stop Training button. - This is a common bug. Close the console to stop RVC entirely. === ==- :question: I couldn't find my answer. - Report your issue [here](https://docs.aihub.gg/contributions). === ###### ‎ :::content-center #### `You have reached the end.` [!badge variant="info" size="xl" corners="pill" icon="paper-airplane" iconAlign="right" text="Report Issues"](https://docs.aihub.gg/contributions/) :::