Home Creators Posts Import Register Favorites Logout

Downloads

Content

FaceFusion 3.5.2 deepfake and face swap app with 1-click installers that now use uv, so installs run up to 100x faster on RunPod and faster on Windows.

The most advanced 1-click deepfake app, installed for FaceFusion V3.5 to use locally on Windows and in the cloud on RunPod and Massed Compute. Works on GPUs from RTX 1000 to RTX 5000 and server GPUs, even on CPU, and reached 60 FPS on an RTX 5090.

FaceFusion 3.5.2 Windows (local), RunPod and Massed Compute (Cloud) 1-Click installers. Install and run with easiness. The best DeepFake APP

Patreon exclusive posts index to find our scripts easily, Patreon scripts updates history to see which updates arrived to which scripts and amazing Patreon special generative scripts list that you can use in any of your task.

Join discord to get help, chat, discuss and also tell me your discord username to get your special rank : SECourses Discord

Please also Star, Watch and Fork our Stable Diffusion & Generative AI  GitHub repository and join our Reddit subreddit and follow me on LinkedIn (my real profile)

=======

Latest installer : FaceFusion_v18.zip

1 January 2026 Update

  • Installers upgraded to uv thus it is 100x faster on RunPod and many times faster on Massed Compute and Windows

  • Installers upgraded to support latest FaceFusion 3.5.2

  • Windows installer doesnt download CUDNN and TensorRT anymore so follow requirements tutorial properly and its updated source post

  • Make sure to follow below requirements tutorial fully and when you start the app you will see like i see at the beginning this

Windows Requirements - Super Important

  • Python 3.10, FFmpeg, CUDA 12.9, C++, CUDNN 9.12, C++ tools, MSVC and Git

    • When you start the app you will see

      • Using CUDA from: C:\Program Files\NVIDIA GPU Computing Toolkit\CUDA\v12.9

      • Using cuDNN from: C:\Program Files\NVIDIA\CUDNN\v9.12\bin\12.9

  • If it doesn't work make sure to follow below tutorial step by step and install everything exactly as shown in this below tutorial

  • https://youtu.be/DrhUHnYfwC0

  • The requirements tutorial post fully updated : https://www.patreon.com/posts/111553210

  • To install on Windows just run Windows_Install.bat

  • To start on Windows just run Windows_Start_App.bat

  • To update on Windows just run Windows_Update.bat

9 September 2025 Update

  • FaceFusion installers are updated for newest FaceFusion version 3.4.0

  • This app became a beast with so many amazing features

  • I have tested latest installer on RunPod, Massed Compute and Windows

  • Now our installer installs with Torch 2.8, CUDA 12.9, cuDNN 9.12 and it is blazing fast

  • If you have older or newer GPU should work on all GPUs starting from RTX 1000 to RTX 5000 or server GPUs

  • Even start with CPU works if you don't have any decent GPU

  • I don't recommend TensorRT since compile takes over 10 min and i found it slower than CUDA but still you can try

  • On RunPod make sure to select template Runpod Pytorch 2.8.0, runpod/pytorch:2.8.0-py3.11-cuda12.8.1-cudnn-devel-ubuntu22.04

  • On Massed Compute make sure to select image Creator > SECourses

  • On Windows we still use Python 3.10.11 and working amazing

  • The installer will auto download cuDNN and TensorRT on Windows but you still need to have accurate CUDA installation and path set as in requirements tutorial

  • On RunPod and Massed Compute just select accurate template and run commands it is fully automatic

  • I recommend a fresh install rather than update, install into a new folder and test

23 May 2025 Update

  • Installers updated for latest FaceFusion 3.3

  • Delete older venv file and run new installer file

  • The installer file download TensorRT_10.9.0.34 libraries and extract into tensorrt_libs folder and now TensorRT will work as well on Windows hopefully

    • TensorRT first time compiles taking huge time at some actions so just wait until screen becomes responsive

  • Massed Compute and RunPod installers now supports RTX 5000 series with Torch 2.7 and CUDA 12.8 - all automatic and supports other older GPUs too

    • Select this template on RunPod :

      • RunPod Pytorch 2.8.0

        runpod/pytorch:2.8.0-py3.11-cuda12.8.1-cudnn-devel-ubuntu22.04

  • The installation and download of models on RunPod can be extremely slow meanwhile Massed Compute is ultra fast (i recommend Massed Compute)

    • Also TensorRT on RunPod or Massed Compute isn't recommended since it takes huge compile time - and i didn't test them - CUDA working ultra fast and perfect

23 March 2025 Update

  • Auto downloaded cuDNN libraries updated to latest 9.8.0

  • TensorRT libraries updated and now works - delete venv and make a fresh install

  • FaceFusion installer upgraded and now supports RTX 5000 series as well

  • Tested on RTX 3090 TI and RTX 5090

  • RTX 5090 reached 60 FPS which is a record

  • Massed Compute installers updated - now faster install and working faster

  • Now installer installs latest version always which is 3.1.1 at the moment

  • FaceFusion is currently the very best FaceSwap / 1-shot Deep Fake app that you can find with so many extra and additional features

  • Do a windows update, then delete venv and then run new installer or make a fresh install

Massed Compute (Recommend Cloud) :

RunPod (Cloud):

Related posts

All posts in this topic: Face Swap, Deepfake and Talking Avatar Apps · New to SECourses? Start Here: Best and Most Popular SECourses Posts

Files

Comments

Mika Kujanpää

I'm installing this to Linux. Should I downgrade triton from 2.2.0 to 2.1.0? Also deepspeed from 0.14.2 to 0.11.2?

Mark Sutherland

I've installed the facefusion but get this error, any help appreciated cheers 2024-05-07 10:13:52.1596399 [E:onnxruntime:Default, provider_bridge_ort.cc:1548 onnxruntime::TryGetProviderInfo_CUDA] D:\a\_work\1\s\onnxruntime\core\session\provider_bridge_ort.cc:1209 onnxruntime::ProviderLibrary::Get [ONNXRuntimeError] : 1 : FAIL : LoadLibrary failed with error 126 "" when trying to load "C:\Users\Mark\Documents\facefusion\facefusion\venv\lib\site-packages\onnxruntime\capi\onnxruntime_providers_cuda.dll" 2024-05-07 10:13:52.3393932 [E:onnxruntime:Default, provider_bridge_ort.cc:1548 onnxruntime::TryGetProviderInfo_CUDA] D:\a\_work\1\s\onnxruntime\core\session\provider_bridge_ort.cc:1209 onnxruntime::ProviderLibrary::Get [ONNXRuntimeError] : 1 : FAIL : LoadLibrary failed with error 126 "" when trying to load "C:\Users\Mark\Documents\facefusion\facefusion\venv\lib\site-packages\onnxruntime\capi\onnxruntime_providers_cuda.dll"

SECourses

i think you dont need should work

SECourses

what is your python version and did you install c++ tools? what is your GPU? reinstall and send me all installation log so i can see if any errors : monstermmorpg@gmail.com i think you didnt follow this tutorial to install as shown there : https://youtu.be/-NjNy7afOQ0

Mark Sutherland

Thanks for the reply. I solved it by just using the cpu. But to answer your question it's on a laptop with a GTX1050i. My PC has a A4000 (16gb) but I have that set up with stable diffusion and one of your one click installs of InstantID and everytime I install something it overwrites the versions etc so I use the laptop for stuff I'm not fussed about. But thanks again for the reply.

SECourses

my installers makes their own virtual environments. so as long as you install them into different folders, they will not affect each other.

Mika Kujanpää

That's correct, and deepspeed dropped some other module version after installing it. I might have made some error because I had plenty of problems with onnxruntime. But I was able to solve them.

droidgame maniac

very informative video, but just use pinokio for facefusion. just works without any problems

Ivan Stoyneshki

I did, I run cuda 11.8 from the tutorial

SECourses

if you become a Gold member i can connect your computer and debug the reason. sadly that message not telling a lot

Adan Aguilar Cisneros

[FACEFUSION.CORE] FFMpeg is not installed i get this error when trying to start the app

Sahil Sangani

2024-06-15 04:25:47.8215414 [E:onnxruntime:Default, provider_bridge_ort.cc:1744 onnxruntime::TryGetProviderInfo_CUDA] C:\a\_work\1\s\onnxruntime\core\session\provider_bridge_ort.cc:1426 onnxruntime::ProviderLibrary::Get [ONNXRuntimeError] : 1 : FAIL : LoadLibrary failed with error 126 "" when trying to load "D:\1Git\face fusion\facefusion\venv\lib\site-packages\onnxruntime\capi\onnxruntime_providers_cuda.dll" *************** EP Error *************** EP Error C:\a\_work\1\s\onnxruntime\python\onnxruntime_pybind_state.cc:866 onnxruntime::python::CreateExecutionProviderInstance CUDA_PATH is set but CUDA wasnt able to be loaded. Please install the correct version of CUDA andcuDNN as mentioned in the GPU requirements page (https://onnxruntime.ai/docs/execution-providers/CUDA-ExecutionProvider.html#requirements), make sure they're in the PATH, and that your GPU is supported. when using [('CUDAExecutionProvider', {'device_id': '0', 'cudnn_conv_algo_search': 'DEFAULT'})] Falling back to ['CUDAExecutionProvider', 'CPUExecutionProvider'] and retrying. **************************************** 2024-06-15 04:25:47.8783911 [E:onnxruntime:Default, provider_bridge_ort.cc:1744 onnxruntime::TryGetProviderInfo_CUDA] C:\a\_work\1\s\onnxruntime\core\session\provider_bridge_ort.cc:1426 onnxruntime::ProviderLibrary::Get [ONNXRuntimeError] : 1 : FAIL : LoadLibrary failed with error 126 "" when trying to load "D:\1Git\face fusion\facefusion\venv\lib\site-packages\onnxruntime\capi\onnxruntime_providers_cuda.dll" Traceback (most recent call last): File "D:\1Git\face fusion\facefusion\venv\lib\site-packages\onnxruntime\capi\onnxruntime_inference_collection.py", line 419, in __init__ self._create_inference_session(providers, provider_options, disabled_optimizers) File "D:\1Git\face fusion\facefusion\venv\lib\site-packages\onnxruntime\capi\onnxruntime_inference_collection.py", line 483, in _create_inference_session sess.initialize_session(providers, provider_options, disabled_optimizers) RuntimeError: C:\a\_work\1\s\onnxruntime\python\onnxruntime_pybind_state.cc:866 onnxruntime::python::CreateExecutionProviderInstance CUDA_PATH is set but CUDA wasnt able to be loaded. Please install the correct version of CUDA andcuDNN as mentioned in the GPU requirements page (https://onnxruntime.ai/docs/execution-providers/CUDA-ExecutionProvider.html#requirements), make sure they're in the PATH, and that your GPU is supported. The above exception was the direct cause of the following exception: Traceback (most recent call last): File "D:\1Git\face fusion\facefusion\run.py", line 6, in core.cli() File "D:\1Git\face fusion\facefusion\facefusion\core.py", line 112, in cli run(program) File "D:\1Git\face fusion\facefusion\facefusion\core.py", line 234, in run ui.launch() File "D:\1Git\face fusion\facefusion\facefusion\uis\core.py", line 88, in launch ui_layout_module.render() File "D:\1Git\face fusion\facefusion\facefusion\uis\layouts\default.py", line 45, in render preview.render() File "D:\1Git\face fusion\facefusion\facefusion\uis\components\preview.py", line 42, in render conditional_append_reference_faces() File "D:\1Git\face fusion\facefusion\facefusion\core.py", line 280, in conditional_append_reference_faces reference_face = get_one_face(reference_frame, facefusion.globals.reference_face_position) File "D:\1Git\face fusion\facefusion\facefusion\face_analyser.py", line 447, in get_one_face many_faces = get_many_faces(vision_frame) File "D:\1Git\face fusion\facefusion\facefusion\face_analyser.py", line 504, in get_many_faces bounding_box_list_yoloface, face_landmark_5_list_yoloface, score_list_yoloface = detect_with_yoloface(vision_frame, facefusion.globals.face_detector_size) File "D:\1Git\face fusion\facefusion\facefusion\face_analyser.py", line 257, in detect_with_yoloface face_detector = get_face_analyser().get('face_detectors').get('yoloface') File "D:\1Git\face fusion\facefusion\facefusion\face_analyser.py", line 95, in get_face_analyser face_detectors['yoloface'] = onnxruntime.InferenceSession(MODELS.get('face_detector_yoloface').get('path'), providers = apply_execution_provider_options(facefusion.globals.execution_device_id, facefusion.globals.execution_providers)) File "D:\1Git\face fusion\facefusion\venv\lib\site-packages\onnxruntime\capi\onnxruntime_inference_collection.py", line 432, in __init__ raise fallback_error from e File "D:\1Git\face fusion\facefusion\venv\lib\site-packages\onnxruntime\capi\onnxruntime_inference_collection.py", line 427, in __init__ self._create_inference_session(self._fallback_providers, None) File "D:\1Git\face fusion\facefusion\venv\lib\site-packages\onnxruntime\capi\onnxruntime_inference_collection.py", line 483, in _create_inference_session sess.initialize_session(providers, provider_options, disabled_optimizers) RuntimeError: C:\a\_work\1\s\onnxruntime\python\onnxruntime_pybind_state.cc:866 onnxruntime::python::CreateExecutionProviderInstance CUDA_PATH is set but CUDA wasnt able to be loaded. Please install the correct version of CUDA andcuDNN as mentioned in the GPU requirements page (https://onnxruntime.ai/docs/execution-providers/CUDA-ExecutionProvider.html#requirements), make sure they're in the PATH, and that your GPU is supported. i have PYTHON 3.10.11. C++ INSTALLED, PYTHON DEVELOPMENT INSTALLED. CUDA 12.1 installed. i added cuDNN to path after downloding it. i installed everything 3 times still same error please help

Sahil Sangani

No error while installation. just this error while using GPU mode. CPU works fine but too slow so i need GPU

SECourses

cuda 12.1 won't work. you have to have cuda 11.8. follow this video and set exactly as in there : https://youtu.be/-NjNy7afOQ0 also install into d:/face_ai folder nothing else. your folder path has space character

Sahil Sangani

I used that before still same. but now i used pinokia installer it worked so thankfully finally. thanks :)

Anshul Gupta

Hi, sir glad to talk with you, you are a genius, I installed face Fusion with your 1 click installer and I am getting the error "run.py: error: argument --execution-providers: invalid choice: 'cuda' (choose from 'CPU')". Please help me to solve it

Anshul Gupta

can you help me and guide me on how to solve the error "cuda out of memory", I think you should make a video on these types of error

Anshul Gupta

Please help on how to reduce the redundancy of large AI models that is large, how to set the path in each model for each AI program and tool, the same models are used by stable diffusion automatic 1111 web UI and stable swarm UI. How to collect all these models in one place and then how to use them for all AI tools to set the path of folders of these models

Anshul Gupta

One last help is needed from you i think Piper TTS is the best voice cloning and Text-to-speech generation AI tool. Can you help me in the cloning of voice as directed by the YouTube channel "Natlamir" in his video "Local Piper TTS Training (Using WSL)" Url link of this video is "https://www.youtube.com/watch?v=kGl0a652XtE&t=1s". Please help me to clone voices in English and hindi. i am ready to pay for it to you. Please Please

SECourses

I will hopefully make a tutorial for the best ai voice cloning. It is not ready yet sadly

SECourses

It depends on each model. You can check their command line arguments to give model folder paths. I show this in my tutorial videos

SECourses

You can't solve it unless the app you are using have optimizations that you didn't enable yet

SECourses

What is your gpu? For Cuda you need Nvidia gpu

Anshul Gupta

i have rtx 4090 paired with i7 14700k

SECourses

it is the best gpu. so my guess would be that you have installation error. please reinstall python c++ exactly as shown here into C drive and uninstall previous ones : https://youtu.be/-NjNy7afOQ0

Blatta

The required onnxruntime 1.17.3 needs CUDA 12.x. From your tutorial i have installed CUDA 11.8. What would be the easiest fix for someone new to windows?

Blatta

Thank you so much for your fast support! I followed the tutorial very closely and was able to get your installs for A1111, Roop unleashed, CodeFormer and LivePortrait_v3 running. But with FaceFusion, when i hunt down the error messages in an internet search, i end up thinking the onnxruntime don´t fit the Cuda 11.8. The requirements.txt points to 11.17.3, the onnxruntime-gpu --upgrade from the bat file leads to a install of 1.18.1. According to https://onnxruntime.ai/docs/execution-providers/CUDA-ExecutionProvider.html this would require Cuda 12.*. But i only understand like 1% of all of this, so i could be completely on a wrong track. You think i should run the steps from the tutorial all again and just give it a new try?

SECourses

I am using Cuda 11.8 in my system and facefusion running. I am using onnxruntime==1.17.1 onnxruntime-gpu==1.17.1

wei smith

can you make 1 click installer for facefusion webcam live?

Steve

Is it me or is this very slow compared to the rope application you posted?

SECourses

yes rope is more optimized and faster. but face fusion is also working for v3 which may speed up

Carlos L. Romero

Can this run local? also flux lora training coming?

SECourses

yes this runs locally. yes hopefully flux training tutorial coming soon

Eyepatch_101

Hi Do you have the Kaggle version ready for this?

SECourses

Kaggle immediately banned my very old account while i was trying to install :/ I still couldn't get it opened yet sadly

SECourses

I am going to add v3 branch installer as well working on it now

Eyepatch_101

Ohh, Thats bad...guess I'll have to wait

SECourses

you have to use paid services or your computer. free service as kaggle and colab won't allow faceswapping they ban the account :/ like use runpod or massed compute

Eyepatch_101

Ohh....okay...out of the two, which one do you recommend as I'm a occasional user (4-5 hrs usage per week) and I only follow your channel...should I go for massed compute or Runpod?

SECourses

massed compute best. 31 cents per hour and much faster to install

Ammar Salem

What is the correct way to update the previews one we had on the 9th of August?

SECourses

for next version please install into a new folder. for regular version you can do just git pull over the folder but there is no update atm

SECourses

i am gonna add windows update button now to zip files

Ammar Salem

Thank you 🙏 This windows update is so important to keep up 😂

SECourses

you are welcome thanks for support

Ammar Salem

When I try to activate AGE MODIFIER I get this error: [FACEFUSION.DOWNLOAD] Validating hash for styleganex_age failed in FaceFusion NEXT

SECourses

i told you to make a fresh install into a new folder :) or you have to keep restarting until it deletes and redownloads all

Ammar Salem

It is in new folder. the only button not working is AGE MODIFIER

Ammar Salem

After restarting it is working fine now. I dont know what was wrong ;-)

SECourses

bunların next diye bir branchı var development branch , 2. zip onu yüklüyor

SECourses

it is mistake of facefusion also it is development branch you can expect it to be broken :D

Jeremy

what would you say is the overall best for deepfake quality this or ROPE?

SECourses

i think rope is the best i don't like facefusion dev :D

SECourses

sorry for late reply. can you make a fresh install and send me all install cmd screen logs? so i can see the install error reason

SECourses

sorry for late reply. can you make a fresh install and send me all install cmd screen logs? so i can see the install error reason

James Adams

Hi, FF Next branch now requires CUDA 12.4.1 or TensorRT 10.3.0. CUDA 11.8 is no longer supported. Can you please update your installer? Thanks!

James Adams

If you could add support for AMD via DirectML, that would be amazing :-)

SECourses

it works already in both cases use latest installer

James Adams

Please add support for AMD! Thank you :-)

SECourses

sadly impossible you need to use cloud

SECourses

it has some advantageous like interface is faster. both are pretty par at the moment

Ammar Salem

I trying using update file and it did not work so I deleted venv and run install again and now i have this error: Enter the execution provider (1 for CUDA - preffered and faster, 2 for CPU): 1 Using CUDA execution provider Traceback (most recent call last): File "F:\face_fusion_v3\facefusion\facefusion.py", line 7, in from facefusion import core File "F:\face_fusion_v3\facefusion\facefusion\core.py", line 8, in from facefusion import content_analyser, face_classifier, face_detector, face_landmarker, face_masker, face_recognizer, logger, process_manager, state_manager, voice_extractor, wording File "F:\face_fusion_v3\facefusion\facefusion\content_analyser.py", line 7, in from facefusion import inference_manager, state_manager, wording File "F:\face_fusion_v3\facefusion\facefusion\inference_manager.py", line 5, in import onnx File "F:\face_fusion_v3\facefusion\venv\lib\site-packages\onnx\__init__.py", line 77, in from onnx.onnx_cpp2py_export import ONNX_ML ImportError: DLL load failed while importing onnx_cpp2py_export: A dynamic link library (DLL) initialization routine failed. Press any key to continue . . .

Paul Pfab

hi, i think i made cuda 12.4 default and i installed LTSC 17.8. getting this: " [E:onnxruntime:Default, provider_bridge_ort.cc:1992 onnxruntime::TryGetProviderInfo_CUDA] D:\a\_work\1\s\onnxruntime\core\session\provider_bridge_ort.cc:1637 onnxruntime::ProviderLibrary::Get [ONNXRuntimeError] : 1 : FAIL : LoadLibrary failed with error 126 "" when trying to load "C:\Users\paul\Facefusion\facefusion\venv\lib\site-packages\onnxruntime\capi\onnxruntime_providers_cuda.dll" nvcc: NVIDIA (R) Cuda compiler driver Copyright (c) 2005-2024 NVIDIA Corporation Built on Tue_Feb_27_16:28:36_Pacific_Standard_Time_2024 Cuda compilation tools, release 12.4, V12.4.99 Build cuda_12.4.r12.4/compiler.33961263_0 any idea? regards paul

SECourses

it is not sufficient our app downloads and extracts dll files. look at the installer source code and verify that those were put into places accurately . face fusion is so annoying to make it run if you also upgrade to gold membership i can connect your pc and try to fix

Paul Pfab

hi, can u provide a link to the last version of your windows installer, that s working with cuda 11.8, please? (before facefusion_3_v1). thx

SECourses

hi it wont work anymore the developer of facefusion deleted older githubs - merged you can manually git checkout to specific commits but may not work if you upgrade gold member i can connect your pc and fix

Error_404_unknown

I had the same error, It's taken me 2 days of install/delete reinstall. Finally I deleted everything py, cuda, VS, ONNX, CDnn and followed the tutorial and now it works. Everything else is now broken but I guess a reinstall of everything will fix that. I needed to clean the junk from my drive anyway this had just expidited the process.

SECourses

awesome i am glad you made it work

Shepard2k

can't get CUDA to work, even with LTSC 17.8 no idea what i'm doing wrong :( [E:onnxruntime:Default, provider_bridge_ort.cc:1992 onnxruntime::TryGetProviderInfo_CUDA] D:\a\_work\1\s\onnxruntime\core\session\provider_bridge_ort.cc:1637 onnxruntime::ProviderLibrary::Get [ONNXRuntimeError] : 1 : FAIL : LoadLibrary failed with error 126 "" when trying to load "C:\Users\Shepard4k\FaceFusion_3_v1\facefusion\venv\lib\site-packages\onnxruntime\capi\onnxruntime_providers_cuda.dll"

SECourses

hopefully making a new video today wait it please - windows 11 fresh install

SECourses

just updated the installers and added info please follow new info and installers

SECourses

just updated the installers and added info please follow new info and installers

SECourses

just updated the installers and added info please follow new info and installers

SECourses

just updated the installers and added info please follow new info and installers

SECourses

just updated the installers and added info please follow new info and installers

ApathyJones

Nice! The newest instructions as of Oct 5 fixed all the Cuda errors for me. The .bat file didn't work because it says it is missing the tqdm package, so I just did the steps manually in lieu of installing it, but regardless, it works now finally :)

Lou

After install and the second step, followed by Start Windows.bat, it just hangs after selecting CUDA execution providers. Downloading:0 percent 0/8kb file=open_nsfw.hash cpu is working at something but nothing happens, just hangs. Did this with the last version too so I gave up. Thought this new version might work.

Albert

I followed the instructions and the red error when generating. 2024-10-05 14:56:26.9995408 [E:onnxruntime:Default, provider_bridge_ort.cc:1992 onnxruntime::TryGetProviderInfo_CUDA] D:\a\_work\1\s\onnxruntime\core\session\provider_bridge_ort.cc:1637 onnxruntime::ProviderLibrary::Get [ONNXRuntimeError] : 1 : FAIL : LoadLibrary failed with error 126 "" when trying to load "D:\FaceFusion_5\facefusion\venv\lib\site-packages\onnxruntime\capi\onnxruntime_providers_cuda.dll"

SECourses

hi you need cudnn fix. verify that dll files are inside D:\FaceFusion_5\facefusion\venv\lib\site-packages\onnxruntime\capi and also this path added to the top of your user path and system path in environment

SECourses

hi it must be your antivirus blocking downloading files. you need to fix it somehow . i would say also try warp vpn

SECourses

ah thanks for reminder i am going to add tqdm package now

SECourses

hi there was an error please try v6 zip file : Windows_Fix_cuDNN_Errors_Run_As_Administrator.bat

Long Phan

I still have the exact same issue, even with the v6 installer. I'll try to reboot

SECourses

then manually verify if all files looking accurate. the capi folder has dlls and enviroment paths are there or not

Long Phan

hmm, the dlls are all there and the environment path has this: (if I execute echo %PATH% from the commande line) \FaceFusion_6\facefusion\venv\Lib\site-packages\onnxruntime\capi; But I still get the same issue.

Arcon Septim

This facefusion completely broke my masedcompute, nothign works

SECourses

it installs some libraries system wide true can break. best to run in a new machine

SECourses

can you show me screenshots of paths of user and system and also screenshot of capi folder and the error you can message from discord

SECourses

i think it is better than roop. but i cant say for sure better than rope pearl - next

Arcon Septim

You have any idea how to remove nsfw censor or any kind of censorship in facefusion?

Lou

Your right. curl.exe was blocked. Thanks.

SECourses

awesome thanks for letting people here know as well

Lou

Alucard24 has a fork of rope with the option to run the development branch. It has features available not in the stable version. Run the development bat install/start in the rope directory to check it out. For me, it runs stable and has the best outputs I've seen so far. I haven't tried Facefusion yet.

SECourses

we have Alucard24 main branch installer and i see both are updated last 2 weeks ago

Lou

Yup, the dev start.bat and dev install.bat should be included in the main rope install folder, from secourses.

jw

facefusion has better eyes and mouth movement adaptation and better colors, but it is way slower than rope , i think due to the use of live_portrait

Error_404_unknown

"You have any idea how to remove nsfw censor or any kind of censorship in facefusion?" I found a way but i don't think i can say here.

SECourses

yes please no NSFW sharing here

bazzamore

Hi Furkan, I noticed in FaceFusion they have option "deep_swapper" which allows users to select a premade face models. Is there a way to train our own models? These seem to produce somewhat better results in some cases.

SECourses

i don't know that. did you find any more info regarding that?

bazzamore

I did look into it a bit further. I believe the training was done via DeepFaceLab then the training is exported to a ".dfm" file.

rino damar jati

can you show me the tutorial to do this?

Michał Kowalski

I think roop is nice for live cam but face fusion gave me better results in quality after swapping, sometimes have to use for example topaz video to make it upscaled, I noticed that facefusion have potential to do it in one move

Roy Davis

Is there a youtube tutorial for this?

PhoB J

Hi, installed all, run Facefusion app bat and throwing: "ValueError: When localhost is not accessible, a shareable link must be created. Please set share=True or check your proxy settings to allow access to localhost. " (it happends the same error also when i install it simply in Pinokio). Can you help?

PhoB J

Thank you for reaction. Im complete noob in python ect. After replacing default.py , now it open web browser(firefox). But shows this: 500 Server Error TypeError: argument of type 'bool' is not iterable Traceback File C:\FaceFusion_11\facefusion\venv\lib\site-packages\gradio_client\utils.py, line 887, in get_type ‒ 884. 885. 886. def get_type(schema: dict): 887. if "const" in schema: 888. return "const" 889. if "enum" in schema: 890. return "enum"

PhoB J

Notieced new v12 installer added, so i used it. Deleted nvec folder , made update and then install. Now it works. Thank you!

PhoB J

But have another puzzle to solve. The tensorrt doesnt work(have rtx5070ti, cuda 12.8.1 installed only). Log says these in red, and continue process using CPU only " 025-03-28 23:13:47.0149933 [E:onnxruntime:Default, provider_bridge_ort.cc:1848 onnxruntime::TryGetProviderInfo_TensorRT] D:\a\_work\1\s\onnxruntime\core\session\provider_bridge_ort.cc:1539 onnxruntime::ProviderLibrary::Get [ONNXRuntimeError] : 1 : FAIL : LoadLibrary failed with error 126 "" when trying to load "c:\FaceFusion\facefusion\venv\lib\site-packages\onnxruntime\capi\onnxruntime_providers_tensorrt.dll" "

SECourses

awesome i made this update to fix it

SECourses

make sure to install cudnn 9.7 . add it to system and user path. after that try again and let me know

PhoB J

Installed 9.7.0 , i do not uninstall the latest 9.8.0 which was installed by you script. In the path, 9.7 is higher than 9.8. But still the same error text.

SECourses

Upgraded V13. now it will download TensorRT libraries into accurate folder. let me know after testing

PhoB J

Yes, you are the man, now works.

PhoB J

Are somewhere any suggestion, what settings use, to squeze most of the framerate? I mean execution thread and queue counts, memory strategy and limit. Im on ryz 9900x, 16gb 5070ti gpu, 64gb ram.

SECourses

sadly i dont have you need to try and compare

decoRyder

OK, wonderful work, as always, and thank you very much. I have CUDA 12.8, and CUDNN 9.8, paths correctly set, and they work in for example ComfyUI. When I run facefusion, I get this: 2025-05-13 09:31:43.3240701 [E:onnxruntime:Default, provider_bridge_ort.cc:2036 onnxruntime::TryGetProviderInfo_CUDA] D:\a\_work\1\s\onnxruntime\core\session\provider_ bridge_ort.cc:1695 onnxruntime::ProviderLibrary::Get [ONNXRuntimeError] : 1 : FAIL : LoadLibrary failed with error 126 "" when trying to load "F:\SECourses\FaceFusion_ 14\facefusion\venv\lib\site-packages\onnxruntime\capi\onnxruntime_providers_cuda.dll I saw that you set the capi path in the script, and no matter if I set the path in my ENV or not I still get the same error. LoadLibrary 126 I think refers to a dendency issue - does facefusion need a specific version of CUDA? Thanks!

koon grizzly

It doesn't want to open the browser by itself when i run the windows start bat file ? C:\audiocraft\audiocraft_plus\venv\lib\site-packages\gradio\components\dropdown.py:231: UserWarning: The value passed into gr.Dropdown() is not in the list of choices. Please update the list of choices to include: No models found or set allow_custom_value=True. warnings.warn( Running on local URL: http://127.0.0.1:7861 So i went to that url myself and it's working fine from there. Quickly tested, is this only photo to photo ? or can it do video like Visomaster? Also it has a lot of censorship, everything nude gets blurred results

SECourses

well this is not developed by me so your best shot is opening issue threads

SECourses

well my downloader downloads and sets into downloaded folder. so you have to use it otherwise really hard to make work

Robert Waters

If you want to bypass the censorship you can replace the content_analyser.py file with: from typing import List # NSFW censorship disabled: all analysis functions now always allow content def pre_check() -> bool: # Skip any model checks return True def clear_inference_pool() -> None: # Stub for FaceFusion pipeline cleanup, no-op pass def analyse_stream(vision_frame, video_fps) -> bool: # Never flag any frame return False def analyse_frame(vision_frame) -> bool: # Never flag any frame return False def analyse_image(image_path: str) -> bool: # Never flag any image return False def analyse_video(video_path: str, trim_frame_start: int, trim_frame_end: int) -> bool: # Never flag any video return False

cool1

With FaceFusion, is there a way to edit a video with someone in but keep the same face but any place where their eyes look down, change it so their eyes look straight ahead instead (facing the camera)? I saw in the info about it that it has "Expression Adjustments". Could that allow that changing eye direction (to straight ahead always)? Or is there something else that could do that better?

SECourses

i dont know you need to play with parameters and test, it supports too many

SECourses

just updated and fixed all. please use this template: RunPod Pytorch 2.8.0 runpod/pytorch:2.8.0-py3.11-cuda12.8.1-cudnn-devel-ubuntu22.04

decoRyder

It straight up doesn't do that, yes you can make rudimentary changes to expressions - happy, angry etc - but you can not change the direction people are facing for example, or where their eyes are looking.

cool1

Thanks a lot. It would have been good if it (or an other AI) could do things like that as someone was asking me to do that for them but it would likely be a lot harder to edit it manually (especially for longer videos) and get it to look realistic enough than if an AI could do it.

Martin Roberts

RunPod PyTorch 2.8.0 container doesn't start, even after setting container disk to 50Gb and 200Gb network storage

SECourses

that means that pod was broken. get a new pod.

Rahul Gautam

Its installing in C drive although I had put the installer in D drive...how Can i change the install location?

Martin Roberts

I couldn't get the PyTorch 2.8 container to run on an RTX 4090 but 5090 was fine

SECourses

dont run as administrator that is your problem

SECourses

yes looks like only starts on rtx 5090 gpu. that is a big issue i will avoid this template in future

Patrick Major

Is this better than the Rope that you have in different sections? I've tried Rope and the UX is quite beautiful (for something that is free). The ability to play and see "live-ish" representations of your changes really helps. Since you have experience with both, which one do you go to most?

SECourses

currently facefusion is most advanced up to date best one

David

Hi, I installed it and in the main root I have TensorRT_10.9.0.34.zip and cudnn_9.12_cuda_12.9.zip, what I have to do?

SECourses

you can delete zip files. it should work now since those are extracted into accurate folders.

David

Ok Ok, thanks!

David Pritchett

It gets stuck when starting : Downloading: 0%| | 0.00/8.00 [00:00

SECourses

try warp vpn . i just tested and no issues. must be related to ISP or antivirus

cawaokaito

Can I install it by simply downloading the FaceFusion folder and double clicking the files inside? Will it download all the necessary installers without having to install Python, git, or conda separately to support rtx5000 series?

SECourses

no you have to have python and git and ffmpeg c++ tools and msvc installed. so first follow this video - 1 time necessary. then simply downloading zip files will work with just 1 click : https://youtu.be/DrhUHnYfwC0

SECourses

I don't know also I don't support NSFW.

ranjeet

I am getting error on windows: Using CUDA from: C:\Program Files\NVIDIA GPU Computing Toolkit\CUDA\v12.4 Using local cuDNN from: D:\AI Apps\FaceFusion_v17_2\cudnn9_12 Checking ONNX Runtime providers... Available providers: ['TensorrtExecutionProvider', 'CUDAExecutionProvider', 'CPUExecutionProvider'] . . . select GPU faster but CPU works too - very slowly Enter the execution provider (1 for CUDA - preffered and faster, 2 for CPU): 1 Using CUDA execution provider C:\Users\ranje\AppData\Local\Programs\Python\Python310\python.exe: can't open file 'D:\\AI Apps\\FaceFusion_v17_2\\facefusion\\facefusion.py': [Errno 2] No such file or directory Press any key to continue . . .

PhoB J

Hi, is it possible to update also cudnn and tensorrt ? Updater and installer install latestest FFversion, but these files start to be dated and i would like to squeeze more from blackwell gpu :) Tried to manualy install newer tensor , but even one revision step not work(i install it by cmd pip install tensorrt==10.10.0.31 --extra-index-url https://pypi.nvidia.com), copied files manually to FF lib+scripts folders, but getting some onnx error when i try tu use it.

SECourses

TensorRT really problematic and also not faster and compile takes forever. increase thread count and compare speeds and set thread count to max speed one. it will make much better impact

PhoB J

I have no delay at initial compile, just several seconds maybe. I had huge delays in the past, but since your V3 script which fixed tensor, it runs without delay. See no big difference(in several frames per sec max) in counts, using 12-16 seems like the best. Just tested just short part of 1080p video, with cuda or tensor , it gets 53-50sec no matter of settings. Almost like tensor not work (but in log there is provider noted as "tensorrt", and GPU load and memory is bit less then when using cuda, in this particular test)

SECourses

wow nice. last time i used this wasnt good

PhoB J

I was able to copy latest 9.17.1 cudnn into cudnn9_12 and capi folder, FF still working. Didnt try to install Cuda 13.x , assuming its general difference, that will mess with everything.

PhoB J

Here error for tensor provider, when i try to install/copy newer one: 01:19:24.1393698 [E:onnxruntime: Default, tensorrt_execution_provider.h:89 onnxruntime::TensorrtLogger::log] [2 025-12-25 00:19:24 ERROR] [stdArchiveReader.cpp::nvinferl::rt::StdArchiveReader::stdArchiveReaderInitCommon::49] Error Code 1: Serialization (Serialization assertion stdVersionRead == kSERIALIZATION_VERSION failed. Version tag does not mat ch. Note: Current Version: 240, Serialized Engine Version: 239 In nvinferl::rt::StdArchiveReader::stdArchiveReaderInitCo mmon at C:\_src\runtime\deserialization\stdArchiveReader.cpp:49) processing: 23%|= 362/1587 [01:15<25:56, 1.27s/frame, execution_providers=['tensorrt']]

PhoB J

or this - the one step newer version 10.10.0.31 - installed, copied, restarted pc. [E:onnxruntime:Default, tensorrt_execution_provider.h:89 onnxruntime::TensorrtLogger::log] [2026-01-01 10:44:36 ERROR] [timingCache.cpp::nvinfer1::builder::TimingCache::validate::949] Error Code 4: API Usage Error (Timing cache header mismatch:Incoming ITimingCache: UUID = GPU-d1159a72-a43c-44eb-0000-000000000000, commit = 9406843d50261530 Runtime device: UUID = GPU-d1159a72-a43c-44eb-00ea-67cb47f163c9, commit = 08bc14bf47a95e54 The incoming cache will not be used!)

SECourses

updated to v18. please read changelogs above

PhoB J

Thank you, you are very productive:) Meanwhile i found some. When i update the tensorrt , first use, even just for a preview, it throws some red errors in log as i posted and program freezes...but i still see little cpu a gpu utilization in task manager...nothing in a log.... when i wait several or more minutes, it finishes and preview is done with detected faces. Then maybe this happed again when i start the job.... but job is done in the end after prolonged period. Maybe sounds like compiling actual job but, since that, every time i use the FF (different videos, after pc restart...) it runs fast with tensor, no errors :) Still on V17 script with tensort 10.13.2

PhoB J

And tested, short part of 1080P video, with CUDA using 35%cpu/85%gpu 13,5gbVram video done after around 145sec. With tensor using 35%cpu/80%gpu a just around 5gbVram i takes around 105sec. Multiple times, every time picked another face reference screen, execution counts 16-20-24-32 doesnt have much influecne, maybe on cuda more counts eats vram to limit(16gb) and start to be even slower(i just skipped worst test). Thank you again and sorry for my noob spamming.

SECourses

yes tensorrt compiles for every new stuff. glad it is working

Christoph Behrmann

E:onnxruntime:Default, provider_bridge_ort.cc:2251 onnxruntime::TryGetProviderInfo_CUDA] E:\_work\1\s\onnxruntime\core\session\provider_bridge_ort.cc:1844 onnxruntime::ProviderLibrary::Get [ONNXRuntimeError] : 1 : FAIL : Error loading "C:\Users\Admin\Downloads\FaceFusion_v18\facefusion\venv\lib\site-packages\onnxruntime\capi\onnxruntime_providers_cuda.dll" which depends on "cudnn64_9.dll" which is missing. Error 126

Stephan Müller

Hallo, unlucky it is not in its own Environment? Right?! So I have to install it over my (lucky) running local system I dont want touch it?! Why?

PhoB J

I see you start using CUDA13 in other software installers, can we somehow do it here? If its worth for blackwell... ?

SECourses

Can be made but software and models has to support otherwise no speed gain. Comfyui benefits since they added relevant code and models

Realtoonai

hi Doc, the title FaceFusion V3.5 but the donwload link is v18, is this the correct one?

SECourses

Yes it is version of the zip file. What it install is the latest facefusion from its repo