SYNTOLOGY HomeExplorerAtlasCodeMethodologyAboutDevelopersFeedPricing
Paper · 2405.20851 · 2024

MegActor: Harness the Power of Raw Video for Vivid Portrait Animation

arXiv · PDF · Open in the Atlas

Code that ran

We lifted 17 functions out of this paper's own repositories and ran 9 of them in a sandbox. "Ran" means the function executed on a synthesized input and returned a value. It is not a reproduction of the paper's results.

RepositoryRoleRan
megvii-research/megfaceanimate canonical 8 of 12
megvii-research/megactor canonical 1 of 5
FunctionStatusWhere it lives
HWC3 Ran megvii-research/megfaceanimate/controlnet_aux_lib/util.py
code served (permissive licence) · get_code("f9ec7d70add02b6f")
get_2d_sincos_pos_embed Ran megvii-research/megactor/animate/megactor-sigma/embeddings.py
code served (permissive licence) · get_code("45f1cb4de58b3f3d")
get_context_scheduler Ran megvii-research/megfaceanimate/animate/context.py
code served (permissive licence) · get_code("d0e5d8aa76eec8a1")
get_total_steps Ran megvii-research/megfaceanimate/animate/context.py
code served (permissive licence) · get_code("859e29c1dc1b42e9")
make_noise_disk Ran megvii-research/megfaceanimate/controlnet_aux_lib/util.py
code served (permissive licence) · get_code("cb55313c2e3ca6a8")
ordered_halving Ran megvii-research/megfaceanimate/animate/context.py
code served (permissive licence) · get_code("ef55c657c8ef7120")
simulate_head_motion Ran megvii-research/megfaceanimate/eval_audio.py
code served (permissive licence) · get_code("ee0d00edd4592e12")
train_collate_fn Ran megvii-research/megfaceanimate/face_dataset.py
code served (permissive licence) · get_code("8b497c21236ea29d")
zero_module Ran megvii-research/megfaceanimate/animate/motion_module.py
code served (permissive licence) · get_code("4719c763c53be3fe")
get_2d_sincos_pos_embed_from_grid Not yet run megvii-research/megactor/animate/megactor-sigma/embeddings.py
code served (permissive licence) · get_code("59fa7b38de05f443")
get_clip_frames Not yet run megvii-research/megfaceanimate/_preprocess.py
code served (permissive licence) · get_code("1db2c40c2d2bd67b")
get_packed_frames Not yet run megvii-research/megfaceanimate/_preprocess.py
code served (permissive licence) · get_code("f350ab376ca5e41c")
get_packed_frames Not yet run megvii-research/megfaceanimate/_preprocess.py
code served (permissive licence) · get_code("1dd5f2925621c29e")
get_timestep_embedding Not yet run megvii-research/megactor/animate/megactor-sigma/embeddings.py
code served (permissive licence) · get_code("38f69a3d1e46016d")
nms Not yet run megvii-research/megfaceanimate/controlnet_aux_lib/util.py
code served (permissive licence) · get_code("6cce57d1077b8c49")
torch_dfs Not yet run megvii-research/megactor/animate/megactor-sigma/stable_diffusion_controlnet_reference.py
code served (permissive licence) · get_code("e55f24470648724c")
zero_module Not yet run megvii-research/megactor/animate/megactor-sigma/controlnet.py
code served (permissive licence) · get_code("da94debb8019ad46")

Repositories linked to this paper

Some links come from the archived Papers with Code dataset (CC BY-SA 4.0): attribution and licence.

Abstract

Despite raw driving videos contain richer information on facial expressions than intermediate representations such as landmarks in the field of portrait animation, they are seldom the subject of research. This is due to two challenges inherent in portrait animation driven with raw videos: 1) significant identity leakage; 2) Irrelevant background and facial details such as wrinkles degrade performance. To harnesses the power of the raw videos for vivid portrait animation, we proposed a pioneering conditional diffusion model named as MegActor. First, we introduced a synthetic data generation framework for creating videos with consistent motion and expressions but inconsistent IDs to mitigate the issue of ID leakage. Second, we segmented the foreground and background of the reference image and employed CLIP to encode the background details. This encoded information is then integrated into the network via a text embedding module, thereby ensuring the stability of the background. Finally, we further style transfer the appearance of the reference image to the driving video to eliminate the influence of facial details in the driving videos. Our final model was trained solely on public datasets, achieving results comparable to commercial models. We hope this will help the open-source community.The code is available at https://github.com/megvii-research/MegFaceAnimate.

For agents

The same record, over MCP at https://syntology.ai/mcp:

get_harvested_code_for_paper("2405.20851")
get_code_for_paper("2405.20851")
have("2405.20851")

Connect an agent — have() is free.