SYNTOLOGY HomeExplorerAtlasCodeMethodologyAboutDevelopersFeedPricing
Paper · 2503.07598 · ICCV · 2025

VACE: All-in-One Video Creation and Editing

arXiv · PDF · Open in the Atlas

Code that ran

11 of 19 samples (functions and classes) lifted from this paper's linked repositories executed on a synthesized input in Syntology's jail (5 of them constructed an object rather than computing a result; 0 honoured the contract we drafted, 1 violated it and 8 have no contract verdict; 2 failures of our instrument; 8 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/2503.07598.json · badge: /api/ran/2503.07598.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
wan-video/wan2.1 community
Papers with Code archive
8 ran · 8 no recorded run
ali-vilab/vace community (classifier label)
classifier basis not recorded on the link
3 ran · 0 no recorded run
FunctionStatusWhere it lives
Head Ran
ran on a synthesized input · constructed an object; no result was computed
wan-video/wan2.1/wan/modules/vace_model.py
code served (permissive licence) · get_code("9227be90bae9b073")
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 an instance of its own class, Head: the run constructed the object and did not call it, so no result was computed. 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 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
import math

torch.manual_seed(0)

# Define the arguments for Head.__init__
dim = 4
out_dim = 2
patch_size = (2, 2)
eps = 1e-06

# Create the Head module
head_module = nn.Module()

# Define ARGS dict with all parameters of Head.__init__
ARGS = {
    'dim': dim,
    'out_dim': out_dim,
    'patch_size': patch_size,
    'eps': eps,
}
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 Head — the run constructed an instance of its own class and did not call it; no shape is stated — the instrument coerces any output to an array; for a Head 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
MLPProj Ran
ran on a synthesized input · constructed an object; no result was computed
wan-video/wan2.1/wan/modules/vace_model.py
code served (permissive licence) · get_code("a30839494a20b1e1")
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 an instance of its own class, MLPProj: the run constructed the object and did not call it, so no result was computed. 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
import torch.nn as nn

torch.manual_seed(0)

# Test case 1: without flf_pos_emb
in_dim_1 = 128
out_dim_1 = 256
flf_pos_emb_1 = False

# Test case 2: with flf_pos_emb
in_dim_2 = 1280
out_dim_2 = 512
flf_pos_emb_2 = True

# We'll use the second case for ARGS since it exercises more code paths
ARGS = {
    'in_dim': in_dim_2,
    'out_dim': out_dim_2,
    'flf_pos_emb': flf_pos_emb_2
}
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 MLPProj — the run constructed an instance of its own class and did not call it; no shape is stated — the instrument coerces any output to an array; for a MLPProj 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
WanLayerNorm Ran
ran on a synthesized input · constructed an object; no result was computed
wan-video/wan2.1/wan/modules/vace_model.py
code served (permissive licence) · get_code("414694f69ae6b8f9")
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 an instance of its own class, WanLayerNorm: the run constructed the object and did not call it, so no result was computed. 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
import torch.nn as nn

torch.manual_seed(0)

ARGS = {
    'dim': 64,
    'eps': 1e-06,
    'elementwise_affine': False,
}
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 WanLayerNorm — the run constructed an instance of its own class and did not call it; no shape is stated — the instrument coerces any output to an array; for a WanLayerNorm 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
WanRMSNorm Ran
ran on a synthesized input · constructed an object; no result was computed
wan-video/wan2.1/wan/modules/vace_model.py
code served (permissive licence) · get_code("f0af233187c10512")
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 an instance of its own class, WanRMSNorm: the run constructed the object and did not call it, so no result was computed. 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
import torch.nn as nn

torch.manual_seed(0)

# WanRMSNorm.__init__ takes: dim, eps=1e-05
# WanRMSNorm.forward takes: x (Tensor of shape [B, L, C])
# The forward method uses x.float(), type_as(x), and self.weight
# _norm uses x.pow(2).mean(dim=-1, keepdim=True)

# Create a small input tensor: [B=2, L=3, C=4]
x = torch.randn(2, 3, 4)

ARGS = {
    "dim": 4,
    "eps": 1e-05,
}
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 WanRMSNorm — the run constructed an instance of its own class and did not call it; no shape is stated — the instrument coerces any output to an array; for a WanRMSNorm 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
WanSelfAttention Ran
ran on a synthesized input · constructed an object; no result was computed
wan-video/wan2.1/wan/modules/vace_model.py
code served (permissive licence) · get_code("f929177cb0050047")
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 an instance of its own class, WanSelfAttention: the run constructed the object and did not call it, so no result was computed. 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 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)

ARGS = {
    'dim': 8,
    'num_heads': 2,
    'window_size': (-1, -1),
    'qk_norm': True,
    'eps': 1e-06,
}
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 WanSelfAttention — the run constructed an instance of its own class and did not call it; no shape is stated — the instrument coerces any output to an array; for a WanSelfAttention 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
filter_args Ran
ran on a synthesized input
ali-vilab/vace/vace/vace_pipeline.py
code served (permissive licence) · get_code("76567eccecdf7882")
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 argparse
import torch

# Create a parser with some test actions
parser = argparse.ArgumentParser(description='Test parser')
parser.add_argument('--learning_rate', type=float, default=0.001)
parser.add_argument('--batch_size', type=int, default=32)
parser.add_argument('--epochs', type=int, default=10)
parser.add_argument('--model_name', type=str, default='bert')

# Create a dict with both known and unknown arguments
args_dict = {
    'learning_rate': 0.001,
    'batch_size': 32,
    'epochs': 10,
    'model_name': 'bert',
    'unknown_param': 'should_be_filtered',
    'another_unknown': 42,
}

ARGS = {
    'args': args_dict,
    '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 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
load_parser Ran
ran on a synthesized input
ali-vilab/vace/vace/vace_pipeline.py
code served (permissive licence) · get_code("eb2897a72245d867")
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 importlib
import sys
from types import ModuleType

# Create a mock module that has a get_parser function
mock_module = ModuleType('mock_parser_module')

def mock_get_parser():
    parser = argparse.ArgumentParser(description='Mock parser')
    parser.add_argument('--name', type=str, default='test')
    parser.add_argument('--value', type=int, default=42)
    return parser

mock_module.get_parser = mock_get_parser

# Register the mock module so importlib can find it
sys.modules['mock_parser_module'] = mock_module

ARGS = {
    'module_name': 'mock_parser_module'
}
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 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
parse_bboxes Ran
ran on a synthesized input
ali-vilab/vace/vace/vace_preproccess.py
code served (permissive licence) · get_code("73fe09ee58b99eef")
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 2, 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 numpy as np

# Analyze the function:
# - parse_bboxes takes a single parameter 's' which is a string
# - The function splits 's' by whitespace to get bbox_str items
# - Each bbox_str is split by comma and converted to floats
# - Each bbox must have exactly 4 coordinates
# - Returns a list of lists (bboxes)

# Valid test case: a string with multiple bounding boxes, each with 4 coordinates
s = "10.5,20.3,100.2,200.1 5.0,15.0,50.5,75.5 0.1,0.2,0.3,0.4"

ARGS = {
    "s": s
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
numpy, torch — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type list, ndim 2, 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
rope_apply Ran
ran; our contract draft was wrong, not the code
wan-video/wan2.1/wan/modules/vace_model.py
code served (permissive licence) · get_code("1d28c4a06c930d63")
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, 10, 4, 12) as a later run recorded it. 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, 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

torch.manual_seed(0)

# Analyze the function:
# - x: indexed as x[i, :seq_len] and x[i, seq_len:], so batch dimension at 0
#      .size(2) and .size(3) accessed, so 4D tensor
#      reshaped to (seq_len, n, -1, 2), so shape is (batch, seq, n, c) where c is even
# - grid_sizes: .tolist() called, iterated over to get (f, h, w) tuples
#              .tolist() suggests it's a tensor
# - freqs: .split() called with 3 parts, indexed [0], [1], [2]
#         each part indexed with [:f], [:h], [:w]
#         so shape is (max_len, c) where c = c - 2*(c//3) + c//3 + c//3

batch_size = 2
n = 4  # spatial dimension from x.size(2)
c = 6  # half of x.size(3), must be divisible by 3 for split
seq_max = 10  # max sequence length for frequency table

# x: shape (batch, seq_max, n, 2*c)
# We'll use different seq_len per batch item
x = torch.randn(batch_size, seq_max, n, 2 * c, dtype=torch.float32)

# grid_sizes: tensor of shape (batch, 3) with (f, h, w) for each batch
# f*h*w should equal seq_len for each batch item
# Batch 0: f=2, h=2, w=2, seq_len=8
# Batch 1: f=2, h=3, w=1, seq_len=6
grid_sizes = torch.tensor([[2, 2, 2], [2, 3, 1]], dtype=torch.long)

# freqs: shape (seq_max, c)
# Will be split into [c - 2*(c//3), c//3, c//3] = [2, 2, 2]
freqs = torch.randn(seq_max, c, dtype=torch.float32)

ARGS = {
    "x": x,
    "grid_sizes": grid_sizes,
    "freqs": freqs,
}
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 Tensor, ndim 4, dtype family float, shape (2, 10, 4, 12) — shape (2, 10, 4, 12) 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
rope_params Ran
ran, violated the contract we drafted
wan-video/wan2.1/wan/modules/vace_model.py
code served (permissive licence) · get_code("dc160d63c5846f4a")
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 2, dtype family float, shape (8, 2) as a later run recorded it. 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, 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.cuda.amp as amp

torch.manual_seed(0)

ARGS = {
    "max_seq_len": 8,
    "dim": 4,
    "theta": 10000
}
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 Tensor, ndim 2, dtype family float, shape (8, 2) — shape (8, 2) 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
sinusoidal_embedding_1d Ran
ran; the contract check could not run (RAISES, DEP_MISSING or TIMEOUT)
wan-video/wan2.1/wan/modules/vace_model.py
code served (permissive licence) · get_code("e34a97853e924d88")
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 2, dtype family float, shape (4, 4) 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

torch.manual_seed(0)

# Analyze the function:
# - dim: int, must be even, used as dimension for embedding
# - position: tensor, used with .type(torch.float64), torch.outer(), and .to() operations
#   position should be 1D tensor of positions

dim = 4  # Must be even
position = torch.tensor([0, 1, 2, 3], dtype=torch.int64)

ARGS = {
    "dim": dim,
    "position": position,
}
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 Tensor, ndim 2, dtype family float, shape (4, 4) — shape (4, 4) 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
BaseWanAttentionBlock Not yet run
unverified (harvested, no recorded run)
wan-video/wan2.1/wan/modules/vace_model.py
code served (permissive licence) · get_code("d8fbfa2bfc03756b")
How it ran

no invocation recorded.

VaceWanAttentionBlock Not yet run
unverified (harvested, no recorded run)
wan-video/wan2.1/wan/modules/vace_model.py
code served (permissive licence) · get_code("3bfd0f1f0b483509")
How it ran

no invocation recorded.

VaceWanModel Not yet run
unverified (harvested, no recorded run)
wan-video/wan2.1/wan/modules/vace_model.py
code served (permissive licence) · get_code("f640e494f36aac4d")
How it ran

no invocation recorded.

WanAttentionBlock Not yet run
unverified (harvested, no recorded run)
wan-video/wan2.1/wan/modules/vace_model.py
code served (permissive licence) · get_code("1e5313e0a5236094")
How it ran

no invocation recorded.

WanI2VCrossAttention Not yet run
unverified (harvested, no recorded run)
wan-video/wan2.1/wan/modules/vace_model.py
code served (permissive licence) · get_code("9a0c2329e53e9f65")
How it ran

no invocation recorded.

WanModel Not yet run
unverified (harvested, no recorded run)
wan-video/wan2.1/wan/modules/vace_model.py
code served (permissive licence) · get_code("cf6067d2fe5bbb8c")
How it ran

no invocation recorded.

WanT2VCrossAttention Not yet run
unverified (harvested, no recorded run)
wan-video/wan2.1/wan/modules/vace_model.py
code served (permissive licence) · get_code("1bae0faad8dabcc7")
How it ran

no invocation recorded.

flash_attention Not yet run
unverified (harvested, no recorded run)
wan-video/wan2.1/wan/modules/vace_model.py
code served (permissive licence) · get_code("7ea44a4680b76539")
How it ran

no invocation recorded.

Also present in other papers' code

Of 19 samples lifted from this paper's linked repositories, 19 carry a function identity and 0 do not. Of the 19 distinct functions with an identity, 4 are also present in other papers whose code we hold (2 with at least one paper dated earlier, 0 the same month, 4 later) and 15 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
sinusoidal_embedding_1d
brick of this paper's assembly: helper
4 other papers
1 dated earlier, 3 later
2502.11079 (2025-02)
Phantom: Subject-Consistent Video Generation via Cross-Modal Alignment
2505.23742 (2025-05)
MAGREF: Masked Guidance for Any-Reference Video Generation with Subject Disentanglement
2507.06119 (2025-07)
Omni-Video: Democratizing Unified Video Understanding and Generation
and 1 more not named here
flash_attention
brick of this paper's assembly: helper
3 other papers
1 dated earlier, 2 later
2502.11079 (2025-02)
Phantom: Subject-Consistent Video Generation via Cross-Modal Alignment
2505.23742 (2025-05)
MAGREF: Masked Guidance for Any-Reference Video Generation with Subject Disentanglement
2604.18215 (2026-04)
Memorize When Needed: Decoupled Memory Control for Spatially Consistent Long-Horizon Video Generation
rope_params
brick of this paper's assembly: helper
2 other papers
2 later
2505.23742 (2025-05)
MAGREF: Masked Guidance for Any-Reference Video Generation with Subject Disentanglement
2604.18215 (2026-04)
Memorize When Needed: Decoupled Memory Control for Spatially Consistent Long-Horizon Video Generation
rope_apply
brick of this paper's assembly: helper
1 other paper
1 later
2505.23742 (2025-05)
MAGREF: Masked Guidance for Any-Reference Video Generation with Subject Disentanglement

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

Diffusion Transformer has demonstrated powerful capability and scalability in generating high-quality images and videos. Further pursuing the unification of generation and editing tasks has yielded significant progress in the domain of image content creation. However, due to the intrinsic demands for consistency across both temporal and spatial dynamics, achieving a unified approach for video synthesis remains challenging. We introduce VACE, which enables users to perform Video tasks within an All-in-one framework for Creation and Editing. These tasks include reference-to-video generation, video-to-video editing, and masked video-to-video editing. Specifically, we effectively * Equal Contribution. † Project lead.

For agents

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

get_harvested_code_for_paper("2503.07598")
get_code_for_paper("2503.07598")
have("2503.07598")

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

curl https://syntology.ai/api/ran/2503.07598.json

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

[![Syntology run record](https://syntology.ai/api/ran/2503.07598.svg)](https://syntology.ai/paper/2503.07598)

Connect an agent — have() is free.