![]() ![]() SRC - always refers to content (frames, faces) of the person whose face we are trying to swap into a target video or photo. workspace - this is where your models, videos, frames, datasets and final video outputs are._internal - internal files, stuff that makes DFL work, No Touchy!.bat, these scripts are used to run various processes required to create a deepfakes, in the main folder you'll see them and 2 folders: for modern AMD cards use "DeepFaceLab_DirectX12" builds (may not work on some older AMD GPUs). for Nvidia RTX 3000 series cards and other GPUs utilizing the same architectures use "DeepFaceLab_NVIDIA_RTX3000_series" builds. for Nvidia GTX 900-1000 and RTX 2000 series and other GPUs utilizing the same architectures as those series use "DeepFaceLab_NVIDIA_up_to_RTX2080Ti" builds. Download correct build of DFL for your GPU (build naming scheme may change): Lastly make sure to enable Hardware-Accelerated GPU Scheduling under Windows 10/11.Ģ. When installing new drivers (be it Studio/Game Ready drivers or AMD drivers) select clean install. When updating drivers consider downloading Display Driver Uninstaller and running it in Safe Mode to fully delete existing drivers, download it from Guru3D: Download DDU You will also need to make sure your OS is updated as well as all your drivers but primarly your GPU drivers, if you are using Nvidia consider switching to Studio Drivers which are evaluated to offer better stability and compatibility over Game Ready Drivers. Power supply - training often can take days and a weak, no-name PSU may fail on you, possibly damaging other hardware, make sure your PSU is from a good brand and isn't an entry level/budget unit, also make sure it meets power requirements of your configuration, use the PSU power calculator below to check if your unit is good enough, if not upgrade it. You may be able to get away with less or auto settings if you have single gpu, for multiple GPU setups I recommend 5x just to be safe.Ĭooling - you will need to make sure your hardware is cooled adequately (check temps of your CPU and GPU during heavy use, if anything is above 85 degrees consider changign your CPU cooler and replacing thermal paste on both CPU and GPU if you know how to do it). (or 5x to be sure if you have less, say 3x3060 = 36GB on 32GB system). plenty of fast SSD storage space and pagefile set to 4x of RAM size minimum if you have more RAM than total of all GPU's VRAM if you plan on using all GPUs at the same time (good for up to 256 resolutions models, for 256-320 11-12GB required, for 320+ 16-24GB GPUs required) 1000 series or higher Nvidia GPU with 8GB VRAM minimum (in general best to have more RAM than total VRAM of all gpus, so 2x4090 = 2x24GB = 48GB, considering modern kits use 16/32GB sticks that would be 64GB of RAM) 32GB of RAM for single GPU configuration, 64GB+ for 2+ GPUs modern 8-32 core CPU supporting AVX and SSE instructions plenty of storage space and pagefile set to 4 x of RAM size minimum. modern Nvidia or AMD GPU with 6GB of VRAM (good for up to 192 resolution models) modern 4 core CPU supporting AVX and SSE instructions Minimum requirements for making very basic and low quality/resolution deepfakes: ![]() Make sure you use standard versions, using N/KN version of Windows may break DFL, if you use N/KN version of Windows install a free Media Feature Pack from Windows Store to ensure DFL works properly. Windows 10 is generally recommended for most users but more advanced users may want to use Linux to get better performance. Usage of Deep Face Lab 2.0 requires high performance PC with modern GPU, ample RAM, storage and fast CPU. However, keep in mind that Google has banned making deepfakes on free accounts, you will have to buy a Colab Pro subsription or higher plan in order to use it (please check TOS before paying, I don't take responsibility if suddenly Google bans use of Colab DFL impletmentation also on paid plans).Ĭolab guide and link to original implementation: ĭFL paper (technical breakdown of the code): If you don't have a powerful PC with GPU with at least 8GB of VRAM you can use Google Colab's DFL implementation to train your models (but rest of the steps will have to be done locally). DFL 2.0 DOWNLOAD (GITHUB, MEGA AND TORRENT): DOWNLOADĭFL 2.0 GITHUB PAGE (new updates, technical support and issues reporting): GITHUB IF YOU LEARNED SOMETHING USEFUL, CONSIDER A DONATION SO I CAN KEEP MAINTAINING THIS GUIDE, IT TOOK MANY HOURS TO WRITE. ![]() IF YOU WANT TO MAKE YOUR OWN GUIDE BASED ON MINE OR ARE REPOSTING IT, PLEASE CREDIT ME, DON'T STEAL IT. READ ENTIRE GUIDE, AS WELL AS FAQS AND USE THE SEARCH OPTION BEFORE YOU POST A NEW QUESTION OR CREATE A NEW THREAD ABOUT AN ISSUE YOU'RE EXPERIENCING! ![]()
0 Comments
Leave a Reply. |
AuthorWrite something about yourself. No need to be fancy, just an overview. ArchivesCategories |