SYNTOLOGY HomeExplorerAtlasCodeMethodologyAboutDevelopersFeedPricing
Paper · 2412.00733 · 2024

Hallo3: Highly Dynamic and Realistic Portrait Image Animation with Video Diffusion Transformer

arXiv · PDF · Open in the Atlas

Code that ran

21 of 23 samples (functions and classes) lifted from this paper's linked repositories executed on a synthesized input in Syntology's jail (0 of them constructed an object rather than computing a result; 3 honoured the contract we drafted, 4 violated it and 0 have no contract verdict; 14 failures of our instrument; 2 no recorded run), measured 2026-11-22; repository links from the Papers with Code archive attributed under CC BY-SA, no affiliation; bodies inline only under a permissive licence as GitHub detects it. Not a reproduction of the paper's results.

"Ran" means the function executed on a synthesized input and returned a value. It is not a reproduction of the paper's results. This record, dated, as JSON: /api/ran/2412.00733.json · badge: /api/ran/2412.00733.svg

Each repository's role is a label: a classifier's output recorded on the link, or, where no role is recorded, what the Papers with Code archive marked. The line under it names its basis. Neither says who wrote the code.

RepositoryRoleRan
fudan-generative-vision/hallo3 canonical (classifier label)
classifier basis: census
21 ran · 2 no recorded run
FunctionStatusWhere it lives
add_model_config_args Ran
ran; our contract draft was wrong, not the code
fudan-generative-vision/hallo3/hallo3/arguments.py
code served (permissive licence) · get_code("ed0fa4fe4bafc350")
How it ran

Ran in Syntology's sandbox on this synthesized input (drafted by a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API) and returned ArgumentParser. The image it ran in was not recorded. If your copy returns the same type on these inputs, it matched what we observed; nothing here checks that. It does NOT mean the output is correct: nothing has compared it to the paper.

import argparse
import torch
import torch.nn as nn

torch.manual_seed(0)

# Create a minimal argument parser
parser = argparse.ArgumentParser()

# Call the function to set up model config arguments
from argparse import ArgumentParser
parser = ArgumentParser()

# Define the ARGS dictionary with values for each parameter
ARGS = {
    'parser': parser
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
torch — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type ArgumentParser — a ArgumentParser output has no array shape of its own; none is stated — the instrument coerces any output to an array; for a ArgumentParser output the recorded ndim 0 and dtype family float describe that coercion, not the output, and are not stated
Environment
image not recorded; re-verify against syntology-harvest-sandbox:v5-deps; not recorded for this sample: the harness installs a cuda-to-cpu shim (cuda_shim.py) before every sample and no run records whether a sample needed it; seed: none by the harness; any seed is in the setup block
add_others_config_args Ran
ran; our contract draft was wrong, not the code
fudan-generative-vision/hallo3/hallo3/arguments.py
code served (permissive licence) · get_code("30f3a95756b04c98")
How it ran

Ran in Syntology's sandbox on this synthesized input (drafted by a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API) and returned MockArgumentParser. The image it ran in was not recorded. If your copy returns the same type on these inputs, it matched what we observed; nothing here checks that. It does NOT mean the output is correct: nothing has compared it to the paper.

import argparse
import torch
import torch.nn as nn

torch.manual_seed(0)

class MockArgumentParser:
    """Minimal mock of argparse.ArgumentParser for testing"""
    def __init__(self):
        self.groups = {}
    
    def add_argument_group(self, name, description):
        group = MockArgumentGroup(name, description)
        self.groups[name] = group
        return group

class MockArgumentGroup:
    """Minimal mock of argument group"""
    def __init__(self, name, description):
        self.name = name
        self.description = description
        self.arguments = {}
    
    def add_argument(self, *args, **kwargs):
        for arg in args:
            self.arguments[arg] = kwargs
        return self

parser = MockArgumentParser()

ARGS = {
    'parser': parser
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
torch — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type MockArgumentParser — a MockArgumentParser output has no array shape of its own; none is stated — the instrument coerces any output to an array; for a MockArgumentParser output the recorded ndim 0 and dtype family float describe that coercion, not the output, and are not stated
Environment
image not recorded; re-verify against syntology-harvest-sandbox:v5-deps; not recorded for this sample: the harness installs a cuda-to-cpu shim (cuda_shim.py) before every sample and no run records whether a sample needed it; seed: none by the harness; any seed is in the setup block
add_sampling_config_args Ran
ran; our contract draft was wrong, not the code
fudan-generative-vision/hallo3/hallo3/arguments.py
code served (permissive licence) · get_code("0e6cd4a28c32de69")
How it ran

Ran in Syntology's sandbox on this synthesized input (drafted by a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API) and returned ArgumentParser. The image it ran in was not recorded. If your copy returns the same type on these inputs, it matched what we observed; nothing here checks that. It does NOT mean the output is correct: nothing has compared it to the paper.

import argparse
import torch
import torch.nn as nn

# Create a minimal argument parser to test the function
parser = argparse.ArgumentParser()

# The function adds arguments to an existing parser
# We need to call the function and capture the resulting parser
def add_sampling_config_args(parser):
    """Sampling configurations"""
    group = parser.add_argument_group('sampling', 'Sampling Configurations')
    group.add_argument('--output-dir', type=str, default='samples')
    group.add_argument('--input-dir', type=str, default=None)
    group.add_argument('--input-type', type=str, default='cli')
    group.add_argument('--input-file', type=str, default='input.txt')
    group.add_argument('--final-size', type=int, default=2048)
    group.add_argument('--sdedit', action='store_true')
    group.add_argument('--grid-num-rows', type=int, default=1)
    group.add_argument('--force-inference', action='store_true')
    group.add_argument('--lcm_steps', type=int, default=None)
    group.add_argument('--sampling-num-frames', type=int, default=32)
    group.add_argument('--sampling-fps', type=int, default=8)
    group.add_argument('--only-save-latents', type=bool, default=False)
    group.add_argument('--only-log-video-latents', type=bool, default=False)
    group.add_argument('--latent-channels', type=int, default=32)
    group.add_argument('--image2video', action='store_true')
    return parser

ARGS = {
    'parser': parser
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
torch — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type ArgumentParser — a ArgumentParser output has no array shape of its own; none is stated — the instrument coerces any output to an array; for a ArgumentParser output the recorded ndim 0 and dtype family float describe that coercion, not the output, and are not stated
Environment
image not recorded; re-verify against syntology-harvest-sandbox:v5-deps; not recorded for this sample: the harness installs a cuda-to-cpu shim (cuda_shim.py) before every sample and no run records whether a sample needed it; seed: none by the harness; any seed is in the setup block
cast_tuple Ran
ran, honoured the contract we drafted
fudan-generative-vision/hallo3/hallo3/vae_modules/cp_enc_dec.py
code served (permissive licence) · get_code("c1b68cdd9e73316a")
How it ran

Ran in Syntology's sandbox on this synthesized input (drafted by a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API) and returned tuple, ndim 1, dtype family int; no output shape is stated. The image it ran in was not recorded. If your copy returns the same type and ndim on these inputs, it matched what we observed; nothing here checks that. It does NOT mean the output is correct: nothing has compared it to the paper.

import torch

torch.manual_seed(0)

ARGS = {
    "t": 42,
    "length": 3
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
torch — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type tuple, ndim 1, dtype family int — a tuple output has no array shape of its own; none is stated
dtype family
int means an integer dtype; float means any other dtype the instrument saw, bool and complex included (fixture_check.py records only these two families)
Environment
image not recorded; re-verify against syntology-harvest-sandbox:v5-deps; not recorded for this sample: the harness installs a cuda-to-cpu shim (cuda_shim.py) before every sample and no run records whether a sample needed it; seed: none by the harness; any seed is in the setup block
collect_video_folder_paths Ran
ran; our contract draft was wrong, not the code
fudan-generative-vision/hallo3/hallo3/extract_meta_info.py
code served (permissive licence) · get_code("317239190c059383")
How it ran

Ran in Syntology's sandbox on this synthesized input (drafted by a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API) and returned list, ndim 1, dtype family float; no output shape is stated. The image it ran in was not recorded. If your copy returns the same type and ndim on these inputs, it matched what we observed; nothing here checks that. It does NOT mean the output is correct: nothing has compared it to the paper.

from pathlib import Path
import tempfile
import os

# Create a temporary directory structure with video folders
temp_dir = tempfile.mkdtemp()
root_path = Path(temp_dir)

# Create some subdirectories to simulate video folders
video_folder_1 = root_path / "video_001"
video_folder_2 = root_path / "video_002"
video_folder_3 = root_path / "video_003"

video_folder_1.mkdir(exist_ok=True)
video_folder_2.mkdir(exist_ok=True)
video_folder_3.mkdir(exist_ok=True)

# Create a file in root to test that only directories are returned
(root_path / "some_file.txt").touch()

ARGS = {
    "root_path": root_path,
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
none outside the standard library — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type list, ndim 1, dtype family float — a list output has no array shape of its own; none is stated
dtype family
int means an integer dtype; float means any other dtype the instrument saw, bool and complex included (fixture_check.py records only these two families)
Environment
image not recorded; re-verify against syntology-harvest-sandbox:v5-deps; not recorded for this sample: the harness installs a cuda-to-cpu shim (cuda_shim.py) before every sample and no run records whether a sample needed it; seed: none by the harness; any seed is in the setup block
construct_meta_info Ran
ran; our contract draft was wrong, not the code
fudan-generative-vision/hallo3/hallo3/extract_meta_info.py
code served (permissive licence) · get_code("3ec64495f4523050")
How it ran

Ran in Syntology's sandbox on this synthesized input (drafted by a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API) and returned dict. The image it ran in was not recorded. If your copy returns the same type on these inputs, it matched what we observed; nothing here checks that. It does NOT mean the output is correct: nothing has compared it to the paper.

import os
from pathlib import Path
import torch
import tempfile

torch.manual_seed(0)

# Create a temporary directory structure to simulate the frames directory layout
temp_dir = tempfile.TemporaryDirectory()
temp_base = Path(temp_dir.name)

# Create the images directory (this is what will be passed as frames_dir_path)
images_dir = temp_base / "images"
images_dir.mkdir(parents=True, exist_ok=True)

# Create all required files by replacing 'images' with the respective names
face_mask_dir = temp_base / "face_mask"
face_mask_dir.mkdir(parents=True, exist_ok=True)
mask_path = str(face_mask_dir) + ".png"
with open(mask_path, "w") as f:
    f.write("")

face_emb_dir = temp_base / "face_emb"
face_emb_dir.mkdir(parents=True, exist_ok=True)
face_emb_path = str(face_emb_dir) + ".pt"
torch.save(torch.randn(2, 4), face_emb_path)

audio_emb_dir = temp_base / "audio_emb"
audio_emb_dir.mkdir(parents=True, exist_ok=True)
audio_emb_path = str(audio_emb_dir) + ".pt"
torch.save(torch.randn(2, 4), audio_emb_path)

videos_dir = temp_base / "videos"
videos_dir.mkdir(parents=True, exist_ok=True)
video_path = str(videos_dir) + ".mp4"
with open(video_path, "w") as f:
    f.write("")

caption_dir = temp_base / "caption"
caption_dir.mkdir(parents=True, exist_ok=True)
caption_path = str(caption_dir) + ".txt"
with open(caption_path, "w", encoding="utf-8") as f:
    f.write("This is a test caption")

ARGS = {
    "frames_dir_path": images_dir,
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
torch — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type dict — a dict output has no array shape of its own; none is stated — the instrument coerces any output to an array; for a dict output the recorded ndim 0 and dtype family float describe that coercion, not the output, and are not stated
Environment
image not recorded; re-verify against syntology-harvest-sandbox:v5-deps; not recorded for this sample: the harness installs a cuda-to-cpu shim (cuda_shim.py) before every sample and no run records whether a sample needed it; seed: none by the harness; any seed is in the setup block
default Ran
ran, violated the contract we drafted
fudan-generative-vision/hallo3/hallo3/vae_modules/attention.py
code served (permissive licence) · get_code("424012cb37b31172")
How it ran

Ran in Syntology's sandbox on this synthesized input (drafted by a language model, Claude Sonnet 4.5, prompt python_fixture_v1) and returned int, ndim 0, dtype family int, shape () as a later run recorded it. The image it ran in was not recorded. If your copy returns the same type, ndim and shape on these inputs, it matched what we observed; nothing here checks that. It does NOT mean the output is correct: nothing has compared it to the paper.

from inspect import isfunction

def exists(val):
    return val is not None

def default(val, d):
    if exists(val):
        return val
    return d() if isfunction(d) else d

# Test case 1: val exists (not None), d is a regular value
ARGS_1 = {
    'val': 42,
    'd': 100
}

# Test case 2: val is None, d is a regular value
ARGS_2 = {
    'val': None,
    'd': 100
}

# Test case 3: val is None, d is a function
def default_func():
    return 200

ARGS_3 = {
    'val': None,
    'd': default_func
}

# Test case 4: val exists, d is a function
ARGS_4 = {
    'val': "existing_value",
    'd': lambda: "default_value"
}

# Use one of the test cases as the final ARGS
ARGS = ARGS_3
Setup block drafted by
a language model, Claude Sonnet 4.5, prompt python_fixture_v1
Imports
none outside the standard library — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type int, ndim 0, dtype family int, shape () — shape () as a later run recorded it
dtype family
int means an integer dtype; float means any other dtype the instrument saw, bool and complex included (fixture_check.py records only these two families)
Environment
image not recorded; re-verify against syntology-harvest-sandbox:v5-deps; not recorded for this sample: the harness installs a cuda-to-cpu shim (cuda_shim.py) before every sample and no run records whether a sample needed it; seed: none by the harness; any seed is in the setup block
divisible_by Ran
ran, violated the contract we drafted
fudan-generative-vision/hallo3/hallo3/vae_modules/cp_enc_dec.py
code served (permissive licence) · get_code("f912c209e9c899ca")
How it ran

Ran in Syntology's sandbox on this synthesized input (drafted by a language model, Claude Haiku 4.5, prompt python_fixture_v1) and returned bool, ndim 0, dtype family float, shape () as a later run recorded it. The image it ran in was not recorded. If your copy returns the same type, ndim and shape on these inputs, it matched what we observed; nothing here checks that. It does NOT mean the output is correct: nothing has compared it to the paper.

import torch
import numpy as np

torch.manual_seed(0)

# The function divisible_by(num, den) checks if num % den == 0
# It takes two numeric arguments and performs modulo operation

ARGS = {
    "num": 10,
    "den": 2
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1
Imports
numpy, torch — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type bool, ndim 0, dtype family float, shape () — shape () as a later run recorded it
dtype family
int means an integer dtype; float means any other dtype the instrument saw, bool and complex included (fixture_check.py records only these two families)
Environment
image not recorded; re-verify against syntology-harvest-sandbox:v5-deps; not recorded for this sample: the harness installs a cuda-to-cpu shim (cuda_shim.py) before every sample and no run records whether a sample needed it; seed: none by the harness; any seed is in the setup block
exists Ran
ran, violated the contract we drafted
fudan-generative-vision/hallo3/hallo3/vae_modules/attention.py
pointer only (licence: MIT); the stored copy get_code serves, from ThomasMrY/VCT, is pointer only (licence: NONE) · get_code("aa5486a3650902d8")
How it ran

The body of this sample is withheld: this paper's link allows the body inline (MIT), but get_code serves the stored copy, first harvested from ThomasMrY/VCT, and GitHub detected no licence in that repository when it was checked. The setup block that called it is withheld with it, because it is written against that body. source_path and origin say where the code lives, to read under the repository's own terms. A run is recorded.

get_2d_sincos_pos_embed Ran
ran, honoured the contract we drafted
fudan-generative-vision/hallo3/hallo3/dit_video_concat.py
code served (permissive licence) · get_code("cdd29f549144d05e")
How it ran

Ran in Syntology's sandbox on this synthesized input (drafted by a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API) and returned ndarray, ndim 2, dtype family float, shape (14, 8) as a later run recorded it. The image it ran in was not recorded. If your copy returns the same type, ndim and shape on these inputs, it matched what we observed; nothing here checks that. It does NOT mean the output is correct: nothing has compared it to the paper.

import numpy as np

# Construct arguments for get_2d_sincos_pos_embed
ARGS = {
    "embed_dim": 8,
    "grid_height": 4,
    "grid_width": 3,
    "cls_token": True,
    "extra_tokens": 2,
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
numpy — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type ndarray, ndim 2, dtype family float, shape (14, 8) — shape (14, 8) as a later run recorded it
dtype family
int means an integer dtype; float means any other dtype the instrument saw, bool and complex included (fixture_check.py records only these two families)
Environment
image not recorded; re-verify against syntology-harvest-sandbox:v5-deps; not recorded for this sample: the harness installs a cuda-to-cpu shim (cuda_shim.py) before every sample and no run records whether a sample needed it; seed: none by the harness; any seed is in the setup block
get_3d_sincos_pos_embed Ran
ran, honoured the contract we drafted
fudan-generative-vision/hallo3/hallo3/dit_video_concat.py
code served (permissive licence) · get_code("7759e9a501ea471a")
How it ran

Ran in Syntology's sandbox on this synthesized input (drafted by a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API) and returned ndarray, ndim 3, dtype family float, shape (2, 12, 16) as a later run recorded it. The image it ran in was not recorded. If your copy returns the same type, ndim and shape on these inputs, it matched what we observed; nothing here checks that. It does NOT mean the output is correct: nothing has compared it to the paper.

import numpy as np

ARGS = {
    "embed_dim": 16,
    "grid_height": 4,
    "grid_width": 3,
    "t_size": 2,
    "cls_token": False,
    "height_interpolation": 1.0,
    "width_interpolation": 1.0,
    "time_interpolation": 1.0,
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
numpy — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type ndarray, ndim 3, dtype family float, shape (2, 12, 16) — shape (2, 12, 16) as a later run recorded it
dtype family
int means an integer dtype; float means any other dtype the instrument saw, bool and complex included (fixture_check.py records only these two families)
Environment
image not recorded; re-verify against syntology-harvest-sandbox:v5-deps; not recorded for this sample: the harness installs a cuda-to-cpu shim (cuda_shim.py) before every sample and no run records whether a sample needed it; seed: none by the harness; any seed is in the setup block
get_batch Ran
ran; the contract check could not run (RAISES, DEP_MISSING or TIMEOUT)
fudan-generative-vision/hallo3/hallo3/sample_video.py
code served (permissive licence) · get_code("4f7e73ff9945218a")
How it ran

Ran in Syntology's sandbox on this synthesized input (drafted by a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API) and returned tuple, ndim 1, dtype family float; no output shape is stated. The image it ran in was not recorded. The code mentions cuda, and the harness aliases cuda to the CPU for every sample; whether this run needed that is not recorded. If your copy returns the same type and ndim on these inputs, it matched what we observed; nothing here checks that. It does NOT mean the output is correct: nothing has compared it to the paper.

import math
from typing import List, Union
from omegaconf import ListConfig
import torch
import numpy as np

torch.manual_seed(0)

ARGS = {
    'keys': ['txt', 'other_key'],
    'value_dict': {
        'prompt': 'a cat sitting on a mat',
        'negative_prompt': 'ugly, blurry',
        'other_key': torch.randn(2, 4),
    },
    'N': [2, 3],
    'T': 16,
    'device': 'cpu',
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
numpy, omegaconf, torch — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type tuple, ndim 1, dtype family float — a tuple output has no array shape of its own; none is stated
dtype family
int means an integer dtype; float means any other dtype the instrument saw, bool and complex included (fixture_check.py records only these two families)
Environment
image not recorded; re-verify against syntology-harvest-sandbox:v5-deps; not recorded for this sample: the harness installs a cuda-to-cpu shim (cuda_shim.py) before every sample and no run records whether a sample needed it; seed: none by the harness; any seed is in the setup block
get_unique_embedder_keys_from_conditioner Ran
ran; our contract draft was wrong, not the code
fudan-generative-vision/hallo3/hallo3/sample_video.py
code served (permissive licence) · get_code("da7df331dab819db")
How it ran

Ran in Syntology's sandbox on this synthesized input (drafted by a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API) and returned list, ndim 1, dtype family float; no output shape is stated. The image it ran in was not recorded. If your copy returns the same type and ndim on these inputs, it matched what we observed; nothing here checks that. It does NOT mean the output is correct: nothing has compared it to the paper.

import torch
import torch.nn as nn

torch.manual_seed(0)

class DummyEmbedder(nn.Module):
    def __init__(self, input_key):
        super().__init__()
        self.input_key = input_key
    
    def forward(self, x):
        return x

class DummyConditioner(nn.Module):
    def __init__(self, embedders):
        super().__init__()
        self.embedders = embedders
    
    def forward(self, x):
        return x

embedder1 = DummyEmbedder("text")
embedder2 = DummyEmbedder("image")
embedder3 = DummyEmbedder("text")

conditioner = DummyConditioner([embedder1, embedder2, embedder3])

ARGS = {
    "conditioner": conditioner
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
torch — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type list, ndim 1, dtype family float — a list output has no array shape of its own; none is stated
dtype family
int means an integer dtype; float means any other dtype the instrument saw, bool and complex included (fixture_check.py records only these two families)
Environment
image not recorded; re-verify against syntology-harvest-sandbox:v5-deps; not recorded for this sample: the harness installs a cuda-to-cpu shim (cuda_shim.py) before every sample and no run records whether a sample needed it; seed: none by the harness; any seed is in the setup block
get_video_paths Ran
ran; our contract draft was wrong, not the code
fudan-generative-vision/hallo3/hallo3/data_preprocess.py
code served (permissive licence) · get_code("b5476e28d6f53521")
How it ran

Ran in Syntology's sandbox on this synthesized input (drafted by a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API) and returned list, ndim 1, dtype family float; no output shape is stated. The image it ran in was not recorded. If your copy returns the same type and ndim on these inputs, it matched what we observed; nothing here checks that. It does NOT mean the output is correct: nothing has compared it to the paper.

import tempfile
from pathlib import Path
import os

# Create a temporary directory with mock video files
temp_dir = tempfile.mkdtemp()
source_dir = Path(temp_dir)

# Create some mock .mp4 files
for i in range(10):
    video_file = source_dir / f"video_{i}.mp4"
    video_file.touch()

# Also create some non-.mp4 files to test filtering
(source_dir / "readme.txt").touch()
(source_dir / "config.json").touch()

ARGS = {
    "source_dir": source_dir,
    "parallelism": 3,
    "rank": 1,
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
none outside the standard library — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type list, ndim 1, dtype family float — a list output has no array shape of its own; none is stated
dtype family
int means an integer dtype; float means any other dtype the instrument saw, bool and complex included (fixture_check.py records only these two families)
Environment
image not recorded; re-verify against syntology-harvest-sandbox:v5-deps; not recorded for this sample: the harness installs a cuda-to-cpu shim (cuda_shim.py) before every sample and no run records whether a sample needed it; seed: none by the harness; any seed is in the setup block
is_odd Ran
ran, violated the contract we drafted
fudan-generative-vision/hallo3/hallo3/vae_modules/cp_enc_dec.py
code served (permissive licence) · get_code("945442cc2f6a19c9")
How it ran

Ran in Syntology's sandbox on this synthesized input (drafted by a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API) and returned bool, ndim 0, dtype family float, shape () as a later run recorded it. The image it ran in was not recorded. If your copy returns the same type, ndim and shape on these inputs, it matched what we observed; nothing here checks that. It does NOT mean the output is correct: nothing has compared it to the paper.

import torch

def divisible_by(num, den):
    return num % den == 0

def is_odd(n):
    return not divisible_by(n, 2)

ARGS = {
    "n": 7
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
torch — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type bool, ndim 0, dtype family float, shape () — shape () as a later run recorded it
dtype family
int means an integer dtype; float means any other dtype the instrument saw, bool and complex included (fixture_check.py records only these two families)
Environment
image not recorded; re-verify against syntology-harvest-sandbox:v5-deps; not recorded for this sample: the harness installs a cuda-to-cpu shim (cuda_shim.py) before every sample and no run records whether a sample needed it; seed: none by the harness; any seed is in the setup block
resize_for_rectangle_crop Ran
ran; the contract check could not run (RAISES, DEP_MISSING or TIMEOUT)
fudan-generative-vision/hallo3/hallo3/data_video.py
code served (permissive licence) · get_code("e0dd6c8c26886b58")
How it ran

Ran in Syntology's sandbox on this synthesized input (drafted by a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API) and returned Tensor, ndim 3, dtype family float, shape (3, 4, 6) as a later run recorded it. The image it ran in was not recorded. If your copy returns the same type, ndim and shape on these inputs, it matched what we observed; nothing here checks that. It does NOT mean the output is correct: nothing has compared it to the paper.

import torch
import numpy as np
import torchvision.transforms as TT
from torchvision.transforms.functional import center_crop, resize
from torchvision.transforms import InterpolationMode

torch.manual_seed(0)
np.random.seed(0)

# Create a small test tensor with shape (batch, channels, height, width)
# Using distinct sizes: batch=1, channels=3, height=5, width=7
arr = torch.randn(1, 3, 5, 7)

# Define image_size as (target_height, target_width)
image_size = (4, 6)

# Define reshape_mode as one of the valid options
reshape_mode = 'random'

ARGS = {
    'arr': arr,
    'image_size': image_size,
    'reshape_mode': reshape_mode
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
numpy, torch, torchvision — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type Tensor, ndim 3, dtype family float, shape (3, 4, 6) — shape (3, 4, 6) as a later run recorded it
dtype family
int means an integer dtype; float means any other dtype the instrument saw, bool and complex included (fixture_check.py records only these two families)
Environment
image not recorded; re-verify against syntology-harvest-sandbox:v5-deps; not recorded for this sample: the harness installs a cuda-to-cpu shim (cuda_shim.py) before every sample and no run records whether a sample needed it; seed: none by the harness; any seed is in the setup block
resize_for_rectangle_crop Ran
ran; the contract check could not run (RAISES, DEP_MISSING or TIMEOUT)
fudan-generative-vision/hallo3/hallo3/sample_video.py
code served (permissive licence) · get_code("5f69a5d32b43f9a1")
How it ran

Ran in Syntology's sandbox on this synthesized input (drafted by a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API) and returned Tensor, ndim 3, dtype family float, shape (3, 4, 5) as a later run recorded it. The image it ran in was not recorded. If your copy returns the same type, ndim and shape on these inputs, it matched what we observed; nothing here checks that. It does NOT mean the output is correct: nothing has compared it to the paper.

import torch
import numpy as np
import torchvision.transforms as TT
from torchvision.transforms.functional import center_crop, resize
from torchvision.transforms import InterpolationMode

torch.manual_seed(0)

# Create a small tensor with shape (batch, channels, height, width)
# Using (1, 3, 6, 8) - batch=1, channels=3, height=6, width=8
arr = torch.randn(1, 3, 6, 8)

# Define image_size as a tuple (height, width)
image_size = (4, 5)

# Define reshape_mode
reshape_mode = 'center'

ARGS = {
    'arr': arr,
    'image_size': image_size,
    'reshape_mode': reshape_mode
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
numpy, torch, torchvision — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type Tensor, ndim 3, dtype family float, shape (3, 4, 5) — shape (3, 4, 5) as a later run recorded it
dtype family
int means an integer dtype; float means any other dtype the instrument saw, bool and complex included (fixture_check.py records only these two families)
Environment
image not recorded; re-verify against syntology-harvest-sandbox:v5-deps; not recorded for this sample: the harness installs a cuda-to-cpu shim (cuda_shim.py) before every sample and no run records whether a sample needed it; seed: none by the harness; any seed is in the setup block
resize_for_square_padding Ran
ran; our contract draft was wrong, not the code
fudan-generative-vision/hallo3/hallo3/data_video.py
code served (permissive licence) · get_code("2482e2916bac95cf")
How it ran

Ran in Syntology's sandbox on this synthesized input (drafted by a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API) and returned Tensor, ndim 4, dtype family float, shape (2, 3, 16, 24) as a later run recorded it. The image it ran in was not recorded. If your copy returns the same type, ndim and shape on these inputs, it matched what we observed; nothing here checks that. It does NOT mean the output is correct: nothing has compared it to the paper.

import torch
import torch.nn.functional as F
from torchvision.transforms.functional import center_crop, resize
from torchvision.transforms import InterpolationMode

torch.manual_seed(0)

# Create a small test tensor with shape (batch, channels, height, width)
# Using distinct dimensions: batch=2, channels=3, height=8, width=8
arr = torch.randn(2, 3, 8, 8)

# image_size should be a tuple/list of two integers
# First element is the target square size, second is the width after padding
image_size = (16, 24)

ARGS = {
    "arr": arr,
    "image_size": image_size,
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
torch, torchvision — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type Tensor, ndim 4, dtype family float, shape (2, 3, 16, 24) — shape (2, 3, 16, 24) as a later run recorded it
dtype family
int means an integer dtype; float means any other dtype the instrument saw, bool and complex included (fixture_check.py records only these two families)
Environment
image not recorded; re-verify against syntology-harvest-sandbox:v5-deps; not recorded for this sample: the harness installs a cuda-to-cpu shim (cuda_shim.py) before every sample and no run records whether a sample needed it; seed: none by the harness; any seed is in the setup block
setup_directories Ran
ran; our contract draft was wrong, not the code
fudan-generative-vision/hallo3/hallo3/data_preprocess.py
code served (permissive licence) · get_code("91eee6d5c860d74a")
How it ran

Ran in Syntology's sandbox on this synthesized input (drafted by a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API) and returned dict. The image it ran in was not recorded. If your copy returns the same type on these inputs, it matched what we observed; nothing here checks that. It does NOT mean the output is correct: nothing has compared it to the paper.

from pathlib import Path
import tempfile
import shutil

# Create a temporary directory structure for testing
temp_base = Path(tempfile.mkdtemp())
video_dir = temp_base / "videos" / "processed"
video_dir.mkdir(parents=True, exist_ok=True)
video_path = video_dir / "test_video.mp4"

# Create the video file (empty file for testing purposes)
video_path.touch()

ARGS = {
    "video_path": video_path
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
none outside the standard library — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type dict — a dict output has no array shape of its own; none is stated — the instrument coerces any output to an array; for a dict output the recorded ndim 0 and dtype family float describe that coercion, not the output, and are not stated
Environment
image not recorded; re-verify against syntology-harvest-sandbox:v5-deps; not recorded for this sample: the harness installs a cuda-to-cpu shim (cuda_shim.py) before every sample and no run records whether a sample needed it; seed: none by the harness; any seed is in the setup block
uniq Ran
ran; our contract draft was wrong, not the code
fudan-generative-vision/hallo3/hallo3/vae_modules/attention.py
pointer only (licence: MIT); the stored copy get_code serves, from JingWu321/MUNBa, is pointer only (licence: NONE) · get_code("9a299fe5ae09e407")
How it ran

The body of this sample is withheld: this paper's link allows the body inline (MIT), but get_code serves the stored copy, first harvested from JingWu321/MUNBa, and GitHub detected no licence in that repository when it was checked. The setup block that called it is withheld with it, because it is written against that body. source_path and origin say where the code lives, to read under the repository's own terms. A run is recorded.

zero_module Ran
ran; our contract draft was wrong, not the code
fudan-generative-vision/hallo3/hallo3/dit_video_concat.py
code served (permissive licence) · get_code("da94debb8019ad46")
How it ran

Ran in Syntology's sandbox on this synthesized input (drafted by a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API) and returned SimpleModule. The image it ran in was not recorded. If your copy returns the same type on these inputs, it matched what we observed; nothing here checks that. It does NOT mean the output is correct: nothing has compared it to the paper.

import torch
from torch import nn

torch.manual_seed(0)

# Create a simple neural network module with some parameters
class SimpleModule(nn.Module):
    def __init__(self):
        super().__init__()
        self.linear1 = nn.Linear(4, 3)
        self.linear2 = nn.Linear(3, 2)
    
    def forward(self, x):
        x = self.linear1(x)
        x = self.linear2(x)
        return x

# Instantiate the module
module = SimpleModule()

# Define the ARGS dictionary
ARGS = {
    "module": module
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
torch — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type SimpleModule — a SimpleModule output has no array shape of its own; none is stated — the instrument coerces any output to an array; for a SimpleModule output the recorded ndim 0 and dtype family float describe that coercion, not the output, and are not stated
Environment
image not recorded; re-verify against syntology-harvest-sandbox:v5-deps; not recorded for this sample: the harness installs a cuda-to-cpu shim (cuda_shim.py) before every sample and no run records whether a sample needed it; seed: none by the harness; any seed is in the setup block
get_2d_sincos_pos_embed_from_grid Not yet run
unverified (harvested, no recorded run)
fudan-generative-vision/hallo3/hallo3/ref_dit_video_concat.py
code served (permissive licence) · get_code("665d8a4e8f673a4c")
How it ran

no invocation recorded.

read_video Not yet run
unverified (harvested, no recorded run)
fudan-generative-vision/hallo3/hallo3/data_video.py
code served (permissive licence) · get_code("ea5ba6b6fc22be49")
How it ran

no invocation recorded.

Also present in other papers' code

Of 23 samples lifted from this paper's linked repositories, 23 carry a function identity and 0 do not. Of the 22 distinct functions with an identity, 15 are also present in other papers whose code we hold (9 with at least one paper dated earlier, 6 the same month, 15 later) and 7 are only here among the 52,977 papers whose code Syntology holds (as of 2026-09-26).

Only here means: among the 52,977 papers whose code Syntology holds (51,887 with lifted samples, 1,090 through their assemblies only), as of 2026-09-26, no other paper's lifted samples or current assembly bricks contain an identical function body after normalisation.

FunctionAlso present inEarliest-datedOthers named
exists 486 other papers
412 dated earlier, 10 the same month, 53 later, 9 same year, order not determinable, 2 undated · 78 only as an assembly brick
1603.08155 (2016-03)
Perceptual Losses for Real-Time Style Transfer and Super-Resolution
1611.07004 (2016-11)
Image-to-Image Translation with Conditional Adversarial Networks
1804.04637 (2018-04)
EMBER: An Open Dataset for Training Static PE Malware Machine Learning Models
and 483 more not named here
default 207 other papers
177 dated earlier, 4 the same month, 20 later, 6 same year, order not determinable · 41 only as an assembly brick
1711.11585 (2017-11)
High-Resolution Image Synthesis and Semantic Manipulation with Conditional GANs
2006.11239 (2020-06)
Denoising Diffusion Probabilistic Models
2010.08162 (2020-10)
SidechainNet: An All-Atom Protein Structure Dataset for Machine Learning
and 204 more not named here
get_2d_sincos_pos_embed_from_grid 190 other papers
133 dated earlier, 6 the same month, 50 later, 1 same year, order not determinable · 65 only as an assembly brick
1807.10221 (2018-07)
Unified Perceptual Parsing for Scene Understanding
2105.12723 (2021-05)
Nested Hierarchical Transformer: Towards Accurate, Data-Efficient and Interpretable Visual Understanding
2111.06377 (2021-11)
Masked Autoencoders Are Scalable Vision Learners
and 187 more not named here
uniq 115 other papers
101 dated earlier, 4 the same month, 6 later, 4 same year, order not determinable
2112.10752 (2021-12)
High-Resolution Image Synthesis with Latent Diffusion Models
2204.06125 (2022-04)
Hierarchical Text-Conditional Image Generation with CLIP Latents
2206.02779 (2022-06)
Blended Latent Diffusion
and 112 more not named here
zero_module 40 other papers
30 dated earlier, 2 the same month, 6 later, 2 same year, order not determinable
2305.13077 (2023-05)
ControlVideo: Training-free Controllable Text-to-Video Generation
2305.17098 (2023-05)
ControlVideo: Conditional Control for One-shot Text-driven Video Editing and Beyond
2308.03448 (2023-08)
Make Explicit Calibration Implicit: Calibrate Denoiser Instead of the Noise Model
and 37 more not named here
divisible_by 16 other papers
9 dated earlier, 7 later · 2 only as an assembly brick
2006.07733 (2020-06)
Bootstrap your own latent: A new approach to self-supervised Learning
2010.10241 (2020-10)
BYOL works even without batch statistics
2102.06810 (2021-02)
Understanding self-supervised Learning Dynamics without Contrastive Pairs
and 13 more not named here
cast_tuple 4 other papers
3 dated earlier, 1 later
2203.07852 (2022-03)
Block-Recurrent Transformers
2303.05511 (2023-03)
Scaling up GANs for Text-to-Image Synthesis
2403.10266 (2024-03)
DSP: Dynamic Sequence Parallelism for Multi-Dimensional Transformers
and 1 more not named here
get_unique_embedder_keys_from_conditioner 4 other papers
2 dated earlier, 1 the same month, 1 later
2405.16947 (2024-05)
Zero-Shot Video Semantic Segmentation based on Pre-Trained Diffusion Models
2406.15333 (2024-06)
GeoLRM: Geometry-Aware Large Reconstruction Model for High-Quality 3D Gaussian Generation
2412.18597 (2024-12)
DiTCtrl: Exploring Attention Control in Multi-Modal Diffusion Transformer for Tuning-Free Multi-Prompt Longer Video Generation
and 1 more not named here
get_2d_sincos_pos_embed 2 other papers
2 later
2502.05179 (2025-02)
FlashVideo: Flowing Fidelity to Detail for Efficient High-Resolution Video Generation
2503.20287 (2025-03)
InsViE-1M: Effective Instruction-based Video Editing with Elaborate Dataset Construction
get_3d_sincos_pos_embed 2 other papers
2 later
2502.05179 (2025-02)
FlashVideo: Flowing Fidelity to Detail for Efficient High-Resolution Video Generation
2503.20287 (2025-03)
InsViE-1M: Effective Instruction-based Video Editing with Elaborate Dataset Construction
get_batch 2 other papers
2 later
2502.05179 (2025-02)
FlashVideo: Flowing Fidelity to Detail for Efficient High-Resolution Video Generation
2503.20287 (2025-03)
InsViE-1M: Effective Instruction-based Video Editing with Elaborate Dataset Construction
is_odd 2 other papers
1 dated earlier, 1 later
2403.10266 (2024-03)
DSP: Dynamic Sequence Parallelism for Multi-Dimensional Transformers
2502.05179 (2025-02)
FlashVideo: Flowing Fidelity to Detail for Efficient High-Resolution Video Generation
add_model_config_args 1 other paper
1 later
2503.20287 (2025-03)
InsViE-1M: Effective Instruction-based Video Editing with Elaborate Dataset Construction
read_video 1 other paper
1 later
2503.20287 (2025-03)
InsViE-1M: Effective Instruction-based Video Editing with Elaborate Dataset Construction
resize_for_rectangle_crop 1 other paper
1 later
2503.20287 (2025-03)
InsViE-1M: Effective Instruction-based Video Editing with Elaborate Dataset Construction

Presence, not direction: an identical function body in two papers' repositories does not say who wrote it, which came from which, or that either paper's method rests on it. Matching is exact after normalisation, so a renamed or edited function is not matched. Dates are when papers appeared, not when code was written. Only .py files the harvester lifted are covered; notebooks are not read. A new-style arXiv id YYMM.NNNN dates a paper to the month 20YY-MM; otherwise Paper.year dates it to the year; otherwise it is undated. Dates are when the paper appeared, not when the code was written. Read 2026-11-22.

Repositories linked to this paper

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

Abstract

Existing methodologies for animating portrait images face significant challenges, particularly in handling non-frontal perspectives, rendering dynamic objects around the portrait, and generating immersive, realistic backgrounds. In this paper, we introduce the first application of a pretrained transformer-based video generative model that demonstrates strong generalization capabilities and generates highly dynamic, realistic videos for portrait animation, effectively addressing these challenges. The adoption of a new video backbone model makes previous U-Net-based methods for identity maintenance, audio conditioning, and video extrapolation inapplicable. To address this limitation, we design an identity reference network consisting of a causal 3D VAE combined with a stacked series of transformer layers, ensuring consistent facial identity across video sequences. Additionally, we investigate various speech audio conditioning and motion frame mechanisms to enable the generation of continuous video driven by speech audio. Our method is validated through experiments on benchmark and newly proposed wild datasets, demonstrating substantial improvements over prior methods in generating realistic portraits characterized by diverse orientations within dynamic and immersive scenes. Further visualizations and the source code are available at: https://fudan-generative-vision.github.io/hallo3/.

For agents

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

get_harvested_code_for_paper("2412.00733")
get_code_for_paper("2412.00733")
have("2412.00733")

The run record, dated, one paper per request, free:

curl https://syntology.ai/api/ran/2412.00733.json

A badge for a README (the split and the date, never a ratio):

[![Syntology run record](https://syntology.ai/api/ran/2412.00733.svg)](https://syntology.ai/paper/2412.00733)

Connect an agent — have() is free.