SYNTOLOGY HomeExplorerAtlasCodeMethodologyAboutDevelopersFeedPricing
Paper · 2312.07537 · 2023

FreeInit: Bridging Initialization Gap in Video Diffusion Models

arXiv · PDF · Open in the Atlas

Code that ran

3 of 3 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; 0 honoured the contract we drafted, 0 violated it and 0 have no contract verdict; 3 failures of our instrument; 0 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/2312.07537.json · badge: /api/ran/2312.07537.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
tianxingwu/freeinit canonical (classifier label)
classifier basis: classifier_2025
3 ran · 0 no recorded run
FunctionStatusWhere it lives
freq_mix_3d Ran
ran; the contract check could not run (RAISES, DEP_MISSING or TIMEOUT)
tianxingwu/freeinit/examples/AnimateDiff/animatediff/utils/freeinit_utils.py
code served (permissive licence) · get_code("e285f4a541803f96")
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 Tensor, ndim 4, dtype family float, shape (2, 4, 3, 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 torch.fft as fft

torch.manual_seed(0)

# Create small 3D tensors for x and noise
# Shape: (batch=2, depth=4, height=3, width=5) to have distinct dimensions
x = torch.randn(2, 4, 3, 5)
noise = torch.randn(2, 4, 3, 5)

# Create a low pass filter with the same spatial dimensions as x and noise
# LPF should be broadcastable to the frequency domain tensors
# Shape: (1, 4, 3, 5) - same as last 3 dims of x/noise
LPF = torch.ones(1, 4, 3, 5) * 0.5  # 0.5 acts as a simple low-pass filter

ARGS = {
    "x": x,
    "noise": noise,
    "LPF": LPF,
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1
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, 4, 3, 5) — shape (2, 4, 3, 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
gaussian_low_pass_filter Ran
ran; our contract draft was wrong, not the code
tianxingwu/freeinit/examples/AnimateDiff/animatediff/utils/freeinit_utils.py
code served (permissive licence) · get_code("5753a1ae1b2b2a7a")
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 Tensor, ndim 5, dtype family float, shape (2, 3, 4, 5, 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 math

torch.manual_seed(0)

# The function expects:
# - shape: a tuple/list representing the shape of a filter volume
# - d_s: normalized stop frequency for spatial dimensions (float, 0.0-1.0)
# - d_t: normalized stop frequency for temporal dimension (float, 0.0-1.0)

# Looking at the function body:
# - It accesses shape[-3], shape[-2], shape[-1] for T, H, W
# - It creates a torch.zeros(shape) tensor
# - It iterates through t, h, w ranges and indexes mask[..., t, h, w]
# - d_s and d_t are used in arithmetic operations

# Create a small 5D shape (batch, channels, time, height, width)
shape = (2, 3, 4, 5, 6)

# Set d_s and d_t to valid normalized frequencies
d_s = 0.25
d_t = 0.25

ARGS = {
    "shape": shape,
    "d_s": d_s,
    "d_t": d_t,
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1
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 5, dtype family float, shape (2, 3, 4, 5, 6) — shape (2, 3, 4, 5, 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
get_freq_filter Ran
ran; the contract check could not run (RAISES, DEP_MISSING or TIMEOUT)
tianxingwu/freeinit/examples/AnimateDiff/animatediff/utils/freeinit_utils.py
code served (permissive licence) · get_code("e6e9076eaee4d40e")
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 Tensor, ndim 5, dtype family float, shape (2, 4, 3, 5, 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 math

torch.manual_seed(0)

# Define arguments for get_freq_filter function
ARGS = {
    'shape': (2, 4, 3, 5, 6),  # (B, C, T, H, W) - batch=2, channels=4, temporal=3, height=5, width=6
    'device': torch.device('cpu'),
    'filter_type': 'gaussian',
    'n': 4,
    'd_s': 0.25,
    'd_t': 0.25,
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1
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 5, dtype family float, shape (2, 4, 3, 5, 6) — shape (2, 4, 3, 5, 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

Also present in other papers' code

Of 3 samples lifted from this paper's linked repositories, 3 carry a function identity and 0 do not. Of the 3 distinct functions with an identity, 3 are also present in other papers whose code we hold (0 with at least one paper dated earlier, 2 the same month, 3 later) and 0 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
gaussian_low_pass_filter 3 other papers
1 the same month, 2 later
2312.16693 (2023-12)
I2V-Adapter: A General Image-to-Video Adapter for Diffusion Models
2402.04324 (2024-02)
ConsistI2V: Enhancing Visual Consistency for Image-to-Video Generation
2501.11319 (2025-01)
StyleSSP: Sampling StartPoint Enhancement for Training-free Diffusion-based Method for Style Transfer
freq_mix_3d 2 other papers
1 the same month, 1 later
2312.16693 (2023-12)
I2V-Adapter: A General Image-to-Video Adapter for Diffusion Models
2402.04324 (2024-02)
ConsistI2V: Enhancing Visual Consistency for Image-to-Video Generation
get_freq_filter 1 other paper
1 later
2402.04324 (2024-02)
ConsistI2V: Enhancing Visual Consistency for Image-to-Video Generation

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

Though diffusion-based video generation has witnessed rapid progress, the inference results of existing models still exhibit unsatisfactory temporal consistency and unnatural dynamics. In this paper, we delve deep into the noise initialization of video diffusion models, and discover an implicit training-inference gap that attributes to the unsatisfactory inference quality.Our key findings are: 1) the spatial-temporal frequency distribution of the initial noise at inference is intrinsically different from that for training, and 2) the denoising process is significantly influenced by the low-frequency components of the initial noise. Motivated by these observations, we propose a concise yet effective inference sampling strategy, FreeInit, which significantly improves temporal consistency of videos generated by diffusion models. Through iteratively refining the spatial-temporal low-frequency components of the initial latent during inference, FreeInit is able to compensate the initialization gap between training and inference, thus effectively improving the subject appearance and temporal consistency of generation results. Extensive experiments demonstrate that FreeInit consistently enhances the generation quality of various text-to-video diffusion models without additional training or fine-tuning.

For agents

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

get_harvested_code_for_paper("2312.07537")
get_code_for_paper("2312.07537")
have("2312.07537")

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

curl https://syntology.ai/api/ran/2312.07537.json

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

[![Syntology run record](https://syntology.ai/api/ran/2312.07537.svg)](https://syntology.ai/paper/2312.07537)

Connect an agent — have() is free.