From b6b6c741a2ce503bcf425f6f95410d2c2d85bc4d Mon Sep 17 00:00:00 2001 From: Kenneth Estanislao Date: Wed, 1 Apr 2026 22:33:01 +0800 Subject: [PATCH] Revert "Merge pull request #1710 from ozp3/amd-dml-optimization" This reverts commit 1b240a45fd58c23a944ca2c9d071a4c6000c2515, reversing changes made to d9a5500bdff12f90511b50b0dd3635a501f44813. --- README.md | 147 ++++++++++------------- modules/core.py | 7 +- modules/face_analyser.py | 10 +- modules/globals.py | 3 - modules/processors/frame/face_swapper.py | 10 +- modules/ui.py | 34 ++---- 6 files changed, 78 insertions(+), 133 deletions(-) diff --git a/README.md b/README.md index ef69748..2fb24dc 100644 --- a/README.md +++ b/README.md @@ -12,7 +12,7 @@ Demo GIF

-## Disclaimer +## Disclaimer This deepfake software is designed to be a productive tool for the AI-generated media industry. It can assist artists in animating custom characters, creating engaging content, and even using models for clothing design. @@ -32,16 +32,14 @@ Users are expected to use this software responsibly and legally. If using a real ## Exclusive v2.7 beta Quick Start - Pre-built (Windows/Mac Silicon/CPU) - + ##### This is the fastest build you can get if you have a discrete NVIDIA or AMD GPU, CPU or Mac Silicon, And you'll receive special priority support. 2.7 beta is the best you can have with 30+ extra features than the open source version. - -###### These Pre-builts are perfect for non-technical users or those who don't have time to, or can't manually install all the requirements. Just a heads-up: this is an open-source project, so you can also install it manually. + +###### These Pre-builts are perfect for non-technical users or those who don't have time to, or can't manually install all the requirements. Just a heads-up: this is an open-source project, so you can also install it manually. ## TLDR; Live Deepfake in just 3 Clicks - ![easysteps](https://github.com/user-attachments/assets/af825228-852c-411b-b787-ffd9aac72fc6) - 1. Select a face 2. Select which camera to use 3. Press live! @@ -111,11 +109,11 @@ This is more likely to work on your computer but will be slower as it utilizes t **1. Set up Your Platform** -- Python (3.11 recommended) -- pip -- git -- [ffmpeg](https://www.youtube.com/watch?v=OlNWCpFdVMA) - `iex (irm ffmpeg.tc.ht)` -- [Visual Studio 2022 Runtimes (Windows)](https://visualstudio.microsoft.com/visual-cpp-build-tools/) +- Python (3.11 recommended) +- pip +- git +- [ffmpeg](https://www.youtube.com/watch?v=OlNWCpFdVMA) - ```iex (irm ffmpeg.tc.ht)``` +- [Visual Studio 2022 Runtimes (Windows)](https://visualstudio.microsoft.com/visual-cpp-build-tools/) **2. Clone the Repository** @@ -127,7 +125,7 @@ cd Deep-Live-Cam **3. Download the Models** 1. [GFPGANv1.4](https://huggingface.co/hacksider/deep-live-cam/resolve/main/GFPGANv1.4.onnx) -2. [inswapper_128_fp16.onnx](https://huggingface.co/hacksider/deep-live-cam/resolve/main/inswapper_128_fp16.onnx) +2. [inswapper\_128\_fp16.onnx](https://huggingface.co/hacksider/deep-live-cam/resolve/main/inswapper_128_fp16.onnx) Place these files in the "**models**" folder. @@ -135,16 +133,14 @@ Place these files in the "**models**" folder. We highly recommend using a `venv` to avoid issues. -For Windows: +For Windows: ```bash python -m venv venv venv\Scripts\activate pip install -r requirements.txt ``` - For Linux: - ```bash # Ensure you use the installed Python 3.10 python3 -m venv venv @@ -158,10 +154,10 @@ Apple Silicon (M1/M2/M3) requires specific setup: ```bash # Install Python 3.11 (specific version is important) -brew install [email protected] +brew install python@3.11 # Install tkinter package (required for the GUI) -brew install [email protected] +brew install python-tk@3.10 # Create and activate virtual environment with Python 3.11 python3.11 -m venv venv @@ -233,20 +229,18 @@ python3.10 run.py --execution-provider coreml ``` **Important Notes for macOS:** - - You **must** use Python 3.10, not newer versions like 3.11 or 3.13 - Always run with `python3.10` command not just `python` if you have multiple Python versions installed -- If you get error about `_tkinter` missing, reinstall the tkinter package: `brew reinstall [email protected]` +- If you get error about `_tkinter` missing, reinstall the tkinter package: `brew reinstall python-tk@3.10` - If you get model loading errors, check that your models are in the correct folder - If you encounter conflicts with other Python versions, consider uninstalling them: - ```bash # List all installed Python versions brew list | grep python - + # Uninstall conflicting versions if needed - brew uninstall --ignore-dependencies [email protected] [email protected] - + brew uninstall --ignore-dependencies python@3.11 python@3.13 + # Keep only Python 3.11 brew cleanup ``` @@ -266,7 +260,7 @@ pip install onnxruntime-coreml==1.21.0 python run.py --execution-provider coreml ``` -**DirectML Execution Provider (Windows AMD/Intel GPU)** +**DirectML Execution Provider (Windows)** 1. Install dependencies: @@ -275,36 +269,12 @@ pip uninstall onnxruntime onnxruntime-directml pip install onnxruntime-directml==1.21.0 ``` -2. Additional packages required: +2. Usage: ```bash -pip install opencv-python customtkinter pillow insightface pygrabber tensorflow "numpy<2" +python run.py --execution-provider directml ``` -3. Usage: - -```bash -python run.py --execution-provider dml -``` - -> **AMD GPU Optimization Notes (RX 5000/6000/7000 series)** -> -> Due to a known `amdxc64.dll` multi-session crash on AMD GPUs, this branch includes the following fixes: -> -> - All models (face analyser + face swapper) are pre-loaded in the main thread before GUI starts to prevent concurrent DML session initialization crashes -> - Face detection is throttled (every 5th frame by default) to maximize GPU swap throughput -> - A global `dml_lock` serializes DML inference calls to prevent race conditions -> - Detection runs inline in the processing thread (detection thread disabled) to avoid concurrent DML access -> -> **Performance on RX 6700 (10GB):** -> -> - No enhancer: ~24 FPS -> - GPEN-256: ~11 FPS -> - GFPGAN: ~8 FPS -> - CPU usage: ~20%, GPU usage: ~60-100% -> -> For quick launch on AMD, use the included `run-dml.bat` file. - **OpenVINO™ Execution Provider (Intel)** 1. Install dependencies: @@ -319,29 +289,27 @@ pip install onnxruntime-openvino==1.21.0 ```bash python run.py --execution-provider openvino ``` - ## Usage **1. Image/Video Mode** -- Execute `python run.py`. -- Choose a source face image and a target image/video. -- Click "Start". -- The output will be saved in a directory named after the target video. +- Execute `python run.py`. +- Choose a source face image and a target image/video. +- Click "Start". +- The output will be saved in a directory named after the target video. **2. Webcam Mode** -- Execute `python run.py`. -- Select a source face image. -- Click "Live". -- Wait for the preview to appear (10-30 seconds). -- Use a screen capture tool like OBS to stream. -- To change the face, select a new source image. +- Execute `python run.py`. +- Select a source face image. +- Click "Live". +- Wait for the preview to appear (10-30 seconds). +- Use a screen capture tool like OBS to stream. +- To change the face, select a new source image. ## Download all models in this huggingface link - - [**Download models here**](https://huggingface.co/hacksider/deep-live-cam/tree/main) ## Command Line Arguments (Unmaintained) @@ -373,32 +341,33 @@ Looking for a CLI mode? Using the -s/--source argument will make the run program ## Press -- [**Ars Technica**](https://arstechnica.com/information-technology/2024/08/new-ai-tool-enables-real-time-face-swapping-on-webcams-raising-fraud-concerns/) - _"Deep-Live-Cam goes viral, allowing anyone to become a digital doppelganger"_ -- [**Yahoo!**](https://www.yahoo.com/tech/ok-viral-ai-live-stream-080041056.html) - _"OK, this viral AI live stream software is truly terrifying"_ -- [**CNN Brasil**](https://www.cnnbrasil.com.br/tecnologia/ia-consegue-clonar-rostos-na-webcam-entenda-funcionamento/) - _"AI can clone faces on webcam; understand how it works"_ -- [**Bloomberg Technoz**](https://www.bloombergtechnoz.com/detail-news/71032/kenalan-dengan-teknologi-deep-live-cam-bisa-jadi-alat-menipu) - _"Get to know Deep Live Cam technology, it can be used as a tool for deception."_ -- [**TrendMicro**](https://www.trendmicro.com/vinfo/gb/security/news/cyber-attacks/ai-vs-ai-deepfakes-and-ekyc) - _"AI vs AI: DeepFakes and eKYC"_ -- [**PetaPixel**](https://petapixel.com/2024/08/14/deep-live-cam-deepfake-ai-tool-lets-you-become-anyone-in-a-video-call-with-single-photo-mark-zuckerberg-jd-vance-elon-musk/) - _"Deepfake AI Tool Lets You Become Anyone in a Video Call With Single Photo"_ -- [**SomeOrdinaryGamers**](https://www.youtube.com/watch?time_continue=1074&v=py4Tc-Y8BcY) - _"That's Crazy, Oh God. That's Fucking Freaky Dude... That's So Wild Dude"_ -- [**IShowSpeed**](https://www.youtube.com/live/mFsCe7AIxq8?feature=shared&t=2686) - _"Alright look look look, now look chat, we can do any face we want to look like chat"_ -- [**TechLinked (Linus Tech Tips)**](https://www.youtube.com/watch?v=wnCghLjqv3s&t=551s) - _"They do a pretty good job matching poses, expression and even the lighting"_ -- [**IShowSpeed**](https://youtu.be/JbUPRmXRUtE?t=3964) - \*"What the F***! Why do I look like Vinny Jr? I look exactly like Vinny Jr!? No, this shit is crazy! Bro This is F*** Crazy!"\* + - [**Ars Technica**](https://arstechnica.com/information-technology/2024/08/new-ai-tool-enables-real-time-face-swapping-on-webcams-raising-fraud-concerns/) - *"Deep-Live-Cam goes viral, allowing anyone to become a digital doppelganger"* + - [**Yahoo!**](https://www.yahoo.com/tech/ok-viral-ai-live-stream-080041056.html) - *"OK, this viral AI live stream software is truly terrifying"* + - [**CNN Brasil**](https://www.cnnbrasil.com.br/tecnologia/ia-consegue-clonar-rostos-na-webcam-entenda-funcionamento/) - *"AI can clone faces on webcam; understand how it works"* + - [**Bloomberg Technoz**](https://www.bloombergtechnoz.com/detail-news/71032/kenalan-dengan-teknologi-deep-live-cam-bisa-jadi-alat-menipu) - *"Get to know Deep Live Cam technology, it can be used as a tool for deception."* + - [**TrendMicro**](https://www.trendmicro.com/vinfo/gb/security/news/cyber-attacks/ai-vs-ai-deepfakes-and-ekyc) - *"AI vs AI: DeepFakes and eKYC"* + - [**PetaPixel**](https://petapixel.com/2024/08/14/deep-live-cam-deepfake-ai-tool-lets-you-become-anyone-in-a-video-call-with-single-photo-mark-zuckerberg-jd-vance-elon-musk/) - *"Deepfake AI Tool Lets You Become Anyone in a Video Call With Single Photo"* + - [**SomeOrdinaryGamers**](https://www.youtube.com/watch?time_continue=1074&v=py4Tc-Y8BcY) - *"That's Crazy, Oh God. That's Fucking Freaky Dude... That's So Wild Dude"* + - [**IShowSpeed**](https://www.youtube.com/live/mFsCe7AIxq8?feature=shared&t=2686) - *"Alright look look look, now look chat, we can do any face we want to look like chat"* + - [**TechLinked (Linus Tech Tips)**](https://www.youtube.com/watch?v=wnCghLjqv3s&t=551s) - *"They do a pretty good job matching poses, expression and even the lighting"* + - [**IShowSpeed**](https://youtu.be/JbUPRmXRUtE?t=3964) - *"What the F***! Why do I look like Vinny Jr? I look exactly like Vinny Jr!? No, this shit is crazy! Bro This is F*** Crazy!"* + ## Credits -- [ffmpeg](https://ffmpeg.org/): for making video-related operations easy -- [Henry](https://github.com/henryruhs): One of the major contributor in this repo -- [deepinsight](https://github.com/deepinsight): for their [insightface](https://github.com/deepinsight/insightface) project which provided a well-made library and models. Please be reminded that the [use of the model is for non-commercial research purposes only](https://github.com/deepinsight/insightface?tab=readme-ov-file#license). -- [havok2-htwo](https://github.com/havok2-htwo): for sharing the code for webcam -- [GosuDRM](https://github.com/GosuDRM): for the open version of roop -- [pereiraroland26](https://github.com/pereiraroland26): Multiple faces support -- [vic4key](https://github.com/vic4key): For supporting/contributing to this project -- [kier007](https://github.com/kier007): for improving the user experience -- [qitianai](https://github.com/qitianai): for multi-lingual support -- [laurigates](https://github.com/laurigates): Decoupling stuffs to make everything faster! -- and [all developers](https://github.com/hacksider/Deep-Live-Cam/graphs/contributors) behind libraries used in this project. -- Footnote: Please be informed that the base author of the code is [s0md3v](https://github.com/s0md3v/roop) -- All the wonderful users who helped make this project go viral by starring the repo ❤️ +- [ffmpeg](https://ffmpeg.org/): for making video-related operations easy +- [Henry](https://github.com/henryruhs): One of the major contributor in this repo +- [deepinsight](https://github.com/deepinsight): for their [insightface](https://github.com/deepinsight/insightface) project which provided a well-made library and models. Please be reminded that the [use of the model is for non-commercial research purposes only](https://github.com/deepinsight/insightface?tab=readme-ov-file#license). +- [havok2-htwo](https://github.com/havok2-htwo): for sharing the code for webcam +- [GosuDRM](https://github.com/GosuDRM): for the open version of roop +- [pereiraroland26](https://github.com/pereiraroland26): Multiple faces support +- [vic4key](https://github.com/vic4key): For supporting/contributing to this project +- [kier007](https://github.com/kier007): for improving the user experience +- [qitianai](https://github.com/qitianai): for multi-lingual support +- [laurigates](https://github.com/laurigates): Decoupling stuffs to make everything faster! +- and [all developers](https://github.com/hacksider/Deep-Live-Cam/graphs/contributors) behind libraries used in this project. +- Footnote: Please be informed that the base author of the code is [s0md3v](https://github.com/s0md3v/roop) +- All the wonderful users who helped make this project go viral by starring the repo ❤️ [![Stargazers](https://reporoster.com/stars/hacksider/Deep-Live-Cam)](https://github.com/hacksider/Deep-Live-Cam/stargazers) @@ -408,4 +377,10 @@ Looking for a CLI mode? Using the -s/--source argument will make the run program ## Stars to the Moon 🚀 - + + + + Star History Chart + + diff --git a/modules/core.py b/modules/core.py index 4280edb..663d742 100644 --- a/modules/core.py +++ b/modules/core.py @@ -2,7 +2,7 @@ import os import sys # single thread doubles cuda performance - needs to be set before torch import if any(arg.startswith('--execution-provider') for arg in sys.argv): - os.environ['OMP_NUM_THREADS'] = '6' + os.environ['OMP_NUM_THREADS'] = '1' # reduce tensorflow log level os.environ['TF_CPP_MIN_LOG_LEVEL'] = '2' import warnings @@ -291,12 +291,9 @@ def run() -> None: for frame_processor in get_frame_processors_modules(modules.globals.frame_processors): if not frame_processor.pre_check(): return - # Pre-load face analyser in main thread before GUI starts - #from modules.face_analyser import get_face_analyser - #get_face_analyser() limit_resources() if modules.globals.headless: start() else: window = ui.init(start, destroy, modules.globals.lang) - window.mainloop() \ No newline at end of file + window.mainloop() diff --git a/modules/face_analyser.py b/modules/face_analyser.py index 5860880..696398b 100644 --- a/modules/face_analyser.py +++ b/modules/face_analyser.py @@ -35,9 +35,7 @@ def get_face_analyser() -> Any: def get_one_face(frame: Frame) -> Any: - import modules.globals as g - with g.dml_lock: - face = get_face_analyser().get(frame) + face = get_face_analyser().get(frame) try: return min(face, key=lambda x: x.bbox[0]) except ValueError: @@ -45,10 +43,8 @@ def get_one_face(frame: Frame) -> Any: def get_many_faces(frame: Frame) -> Any: - import modules.globals as g try: - with g.dml_lock: - return get_face_analyser().get(frame) + return get_face_analyser().get(frame) except IndexError: return None @@ -200,4 +196,4 @@ def dump_faces(centroids: Any, frame_face_embeddings: list): if temp_frame[int(y_min):int(y_max), int(x_min):int(x_max)].size > 0: cv2.imwrite(temp_directory_path + f"/{i}/{frame['frame']}_{j}.png", temp_frame[int(y_min):int(y_max), int(x_min):int(x_max)]) - j += 1 + j += 1 \ No newline at end of file diff --git a/modules/globals.py b/modules/globals.py index 3d88931..aabc19a 100644 --- a/modules/globals.py +++ b/modules/globals.py @@ -71,6 +71,3 @@ interpolation_weight: float = 0 # Blend weight for current frame (0.0-1.0). Low # --- END: Added for Frame Interpolation --- # --- END OF FILE globals.py --- - -import threading -dml_lock = threading.Lock() diff --git a/modules/processors/frame/face_swapper.py b/modules/processors/frame/face_swapper.py index 57bf508..04f846b 100644 --- a/modules/processors/frame/face_swapper.py +++ b/modules/processors/frame/face_swapper.py @@ -110,6 +110,7 @@ def get_face_swapper() -> Any: )) else: providers_config.append(p) + FACE_SWAPPER = insightface.model_zoo.get_model( model_path, providers=providers_config, @@ -152,10 +153,9 @@ def swap_face(source_face: Face, target_face: Face, temp_frame: Frame) -> Frame: if not temp_frame.flags['C_CONTIGUOUS']: temp_frame = np.ascontiguousarray(temp_frame) - with modules.globals.dml_lock: - swapped_frame_raw = face_swapper.get( - temp_frame, target_face, source_face, paste_back=True - ) + swapped_frame_raw = face_swapper.get( + temp_frame, target_face, source_face, paste_back=True + ) # --- START: CRITICAL FIX FOR ORT 1.17 --- # Check the output type and range from the model @@ -1183,4 +1183,4 @@ def apply_color_transfer(source, target): # traceback.print_exc() return source - return result_bgr + return result_bgr \ No newline at end of file diff --git a/modules/ui.py b/modules/ui.py index 6041fc4..7143076 100644 --- a/modules/ui.py +++ b/modules/ui.py @@ -72,8 +72,8 @@ ROOT_WIDTH = 600 PREVIEW = None PREVIEW_MAX_HEIGHT = 700 PREVIEW_MAX_WIDTH = 1200 -PREVIEW_DEFAULT_WIDTH = 640 -PREVIEW_DEFAULT_HEIGHT = 360 +PREVIEW_DEFAULT_WIDTH = 960 +PREVIEW_DEFAULT_HEIGHT = 540 POPUP_WIDTH = 750 POPUP_HEIGHT = 810 @@ -1000,10 +1000,6 @@ def webcam_preview(root: ctk.CTk, camera_index: int): if modules.globals.source_path is None: update_status("Please select a source image first") return - from modules.processors.frame.face_swapper import get_face_swapper - from modules.face_analyser import get_face_analyser - get_face_analyser() - get_face_swapper() create_webcam_preview(camera_index) else: modules.globals.source_target_map = [] @@ -1109,7 +1105,7 @@ def _detection_thread_func(latest_frame_holder, detection_result, detection_lock frame = latest_frame_holder[0] if frame is None: - time.sleep(0.2) + time.sleep(0.005) continue if modules.globals.many_faces: @@ -1161,22 +1157,7 @@ def _processing_thread_func(capture_queue, processed_queue, stop_event, source_image = get_one_face(cv2.imread(modules.globals.source_path)) # Read latest detection results (brief lock to avoid blocking detection thread) - # Run detection inline since detection thread is disabled - # Run detection every 3 frames, reuse cached result otherwise - if not hasattr(_processing_thread_func, '_det_count'): - _processing_thread_func._det_count = 0 - _processing_thread_func._det_count += 1 - - if _processing_thread_func._det_count % 3 == 0: - if modules.globals.many_faces: - cached_target_face = None - cached_many_faces = get_many_faces(temp_frame) - detection_result['many_faces'] = cached_many_faces - else: - cached_target_face = get_one_face(temp_frame) - cached_many_faces = None - detection_result['target_face'] = cached_target_face - else: + with detection_lock: cached_target_face = detection_result.get('target_face') cached_many_faces = detection_result.get('many_faces') @@ -1294,7 +1275,7 @@ def create_webcam_preview(camera_index: int): args=(latest_frame_holder, detection_result, detection_lock, stop_event), daemon=True, ) - # det_thread.start() + det_thread.start() # Start processing thread proc_thread = threading.Thread( @@ -1335,7 +1316,7 @@ def create_webcam_preview(camera_index: int): temp_frame = fit_image_to_size( temp_frame, PREVIEW.winfo_width(), PREVIEW.winfo_height() ) - temp_frame = temp_frame.copy() + image = gpu_cvt_color(temp_frame, cv2.COLOR_BGR2RGB) image = Image.fromarray(image) image = ImageOps.contain( @@ -1593,5 +1574,4 @@ def update_webcam_target( target_label_dict_live[button_num] = target_image else: update_pop_live_status("Face could not be detected in last upload!") - return map - + return map \ No newline at end of file