| Instructions (directions) on how to use AI voice model(s): | |
| 1. If you have RVC installed (WebUI and/or GUI), then no worries - you're almost all set. If not, then you may use Google Colab, but please take note that your account is Pro, since Gradio UI has banned for all who are free users (both Model Inference / Machine Learning and Training) that if you tried and stayed for a long while like 2 or more minutes, you'll be disconnected automatically. | |
| On the other hand, for those who are looking for the alternate training method, you may try these 2 not-yet-restricted links in Colab (both for free and Pro users): | |
| a. https://colab.research.google.com/drive/1PorBmm4bUjY1-dk4f7XKrO7Mx51wbqYU (Training only) | |
| b. https://colab.research.google.com/drive/1r4IRL0UA7JEoZ0ZK8PKfMyTIBHKpyhcw (With model inference feature) | |
| Different versions of RVC can be found here: https://huggingface.co/lj1995/VoiceConversionWebUI/tree/main | |
| *All with the mentioned URLs are required to have internet connection for data processing. | |
| 2. Normally, most authors put their finalized voice models on a .zip file (using WinRAR or 7-Zip) after their own analytical tests had been finished and satisfied from the results they have made. It only contains 2 files: a .pth file & an .index file - both of them are important to work properly. When you make a voice model, then it should be your main basis no more or less. | |
| 3. Extract the .zip file into your own directory of choice. | |
| 4. Now for the 2 important files after you extract it: | |
| >> The .pth file should put in the "weights" folder (both in RVC and for outside sources like Colab). | |
| >> While the .index file should be in the "logs" folder (again, both in RVC and for outside sources like Colab). | |
| >>> If you don't have RVC yet, then put the 2 files somewhere you know - and you can use wherever you need (like the 2 Colab URLs mentioned for you). | |
| 5. Then what's next? I recommend you to open RVC WebUI via "go-web.bat" (double-click) instead of using "go-realtime-gui.bat", for easy convenience. | |
| 6. While waiting to finish loading the WebUI, please don't close the console as that's fully required to run properly. Otherwise, the "Connection errored out" will appear if you intent (or by accident) to close it, and you'll have to restart the kernel again. Unless you are done everything on the WebUI, then you may close. | |
| 7. Do your own experiments on your voice model(s) and enjoy! | |
| ///////==========/////// | |
| WARNING: | |
| I am pretty sure that almost everyone in the world are excited to use any voice models, similar to other AI things by human innovation as well nowadays. But be warned, there's some "sneaky & risky" destructive criticisms about this feature that it is particularly, a "deepfake-esque" work even though this is one of the scientific art pieces (or you may like to say - a crafted masterpiece) to be suggestive enough. Also to mention for the legal issues, "questionable ethics", misuse & abuse, other criminal intents (e.g. blackmail, scam, fraud, "revenge acts", etc.), "religious" immoralities and other negative factors that contributed through this. Plus, doing this so, when you just really love this kind of "freelance job" and you would thought that there won't have any "unforeseen consequences", then that probably likely would be the case of "newly made addiciton" for some or most people. | |
| Unfortunately, this and other AI stuff won't ever stop creating and the future might be uncertain to everyone's knowledge or guess, within the next months, years or decades. Certainly, | |
| As an author myself, why am I actually doing this - is for "vocal (and historical) preservation" within democratic and scientific purposes that I'm taking risks as well currently. I might delete some voice models if there were any kind of complaints, or probably not, depending on their concerns and my "feelings". I will not be liable when something goes wrong in your life, or in other person's life - when relating to this specifically. You will be auto-liable for yourself once you have started anything related to this job, and hopefully there won't be any lawsuits and suing you in the long run, as long as you're fully aware even for myself too. | |
| So in other words, you must be fully responsible, think carefully whatever there are things that are right or wrong, and remain vigilant that you must know what you are doing in your whole life so far. Otherwise well, if you think that making AI voice model (and using model inference for whatever audios you have,) is immoral and rightfully wrong in the first place, then please don't (delete then "forget" them as you like, if you have your own audio datasets and whatever) - simple as that; you have your own choice to decide. When there's a will - there's a way but if there's something you cannot do - there will be always reason(s) behind. | |
| This is my wise advice(s) for anyone who's still reading this "mini-manual". | |
| - TAS Tomusan (Thomas Andre Saldaña) | |
| This file was created in November 22, 2023 (11:27 AM - Oman time) | |
| Last updated: December 2, 2023 (4:48 PM - Oman time) |