SYNTOLOGY HomeExplorerAtlasCodeMethodologyAboutDevelopersFeedPricing
Paper · 2005.08209 · 2020

Learning Individual Speaking Styles for Accurate Lip to Speech Synthesis

arXiv · PDF · Open in the Atlas

Code that ran

8 of 9 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 7 have no contract verdict; 1 failure of our instrument; 1 no recorded run), measured 2026-11-21; 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/2005.08209.json · badge: /api/ran/2005.08209.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
Rudrabha/Lip2Wav canonical (classifier label)
classifier basis: census
8 ran · 1 no recorded run
FunctionStatusWhere it lives
conv3x3 Ran
ran; our contract draft was wrong, not the code
Rudrabha/Lip2Wav/face_detection/models.py
pointer only (licence: MIT); the stored copy get_code serves, from xiexh20/CHORE, is pointer only (licence: NONE) · get_code("5baaa8c1b148ef70")
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 xiexh20/CHORE, 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.

crop Ran
ran on a synthesized input
Rudrabha/Lip2Wav/face_detection/utils.py
code served (permissive licence) · get_code("535d1e92aecfff67")
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 int, shape (256, 256, 3) 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 cv2

torch.manual_seed(0)

# Create a sample RGB image (height=512, width=512, channels=3)
image = np.random.randint(0, 256, size=(512, 512, 3), dtype=np.uint8)

# Center of the object
center = np.array([256.0, 256.0])

# Scale of the face/object
scale = 1.0

# Resolution parameter
resolution = 256.0

ARGS = {
    "image": image,
    "center": center,
    "scale": scale,
    "resolution": resolution
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
cv2, numpy, torch — 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 int, shape (256, 256, 3) — shape (256, 256, 3) 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
crop_frame Ran
ran on a synthesized input
Rudrabha/Lip2Wav/preprocess.py
code served (permissive licence) · get_code("a5424d7dbfd05689")
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 int, shape (190, 360, 3) 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

# Create a minimal frame (numpy array simulating a video frame)
# Using shape (480, 1280, 3) to simulate a typical video frame (height, width, channels)
frame = np.random.randint(0, 256, size=(480, 1280, 3), dtype=np.uint8)

# Create a minimal args object with speaker parameter
class Args:
    def __init__(self, speaker):
        self.speaker = speaker

# Test with 'chess' speaker to exercise the cropping logic
args = Args(speaker='chess')

ARGS = {
    'frame': frame,
    'args': args
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
cv2, 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 int, shape (190, 360, 3) — shape (190, 360, 3) 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_image_list Ran
ran on a synthesized input
Rudrabha/Lip2Wav/complete_test_generate.py
code served (permissive licence) · get_code("8c52e380997759b5")
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 os
import tempfile
import shutil

# Create a temporary directory structure that mimics the expected data layout
temp_dir = tempfile.mkdtemp()

# Create the directory structure
preprocessed_dir = os.path.join(temp_dir, 'preprocessed')
os.makedirs(preprocessed_dir, exist_ok=True)

# Create a sample video ID directory with subdirectories and jpg files
vid_id_1 = 'video_001'
vid_id_2 = 'video_002'

for vid_id in [vid_id_1, vid_id_2]:
    vid_path = os.path.join(preprocessed_dir, vid_id)
    for subdir in ['frame_00', 'frame_01']:
        subdir_path = os.path.join(vid_path, subdir)
        os.makedirs(subdir_path, exist_ok=True)
        # Create dummy jpg files
        for i in range(2):
            jpg_path = os.path.join(subdir_path, f'image_{i}.jpg')
            with open(jpg_path, 'w') as f:
                f.write('dummy jpg content')

# Create the split text file
split_name = 'train'
split_file = os.path.join(temp_dir, f'{split_name}.txt')
with open(split_file, 'w') as f:
    f.write(f'{vid_id_1}\n')
    f.write(f'{vid_id_2}\n')

ARGS = {
    'split': split_name,
    'data_root': temp_dir
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
cv2 — 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_image_list Ran
ran on a synthesized input
Rudrabha/Lip2Wav/synthesizer/hparams.py
code served (permissive licence) · get_code("e69028e4a38f513d")
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
import os

# Create a temporary directory structure that the function expects
temp_dir = tempfile.mkdtemp()

# Create subdirectories for preprocessed data
split_name = "train"
vid_id = "video_001"
preprocessed_dir = os.path.join(temp_dir, "preprocessed", vid_id, "frame_batch_1")
os.makedirs(preprocessed_dir, exist_ok=True)

# Create a dummy jpg file
dummy_jpg = os.path.join(preprocessed_dir, "frame_001.jpg")
with open(dummy_jpg, 'w') as f:
    f.write("dummy")

# Create another frame in a different subdirectory
preprocessed_dir_2 = os.path.join(temp_dir, "preprocessed", vid_id, "frame_batch_2")
os.makedirs(preprocessed_dir_2, exist_ok=True)
dummy_jpg_2 = os.path.join(preprocessed_dir_2, "frame_002.jpg")
with open(dummy_jpg_2, 'w') as f:
    f.write("dummy")

# Create the split file (e.g., train.txt)
split_file = os.path.join(temp_dir, f"{split_name}.txt")
with open(split_file, 'w') as f:
    f.write(f"{vid_id}\n")

ARGS = {
    "split": split_name,
    "data_root": temp_dir,
}
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
inv_preemphasis Ran
ran on a synthesized input
Rudrabha/Lip2Wav/synthesizer/audio.py
code served (permissive licence) · get_code("bd89c559f716b290")
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 1, dtype family float, shape (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
from scipy import signal

# Set seed for reproducibility
torch.manual_seed(0)

# Create a small audio waveform (wav parameter)
# Using numpy array as scipy.signal.lfilter works with array-like inputs
wav = np.array([0.1, 0.2, -0.15, 0.05, 0.3, -0.1], dtype=np.float32)

# Preemphasis coefficient (k parameter)
# Typical values range from 0.95 to 0.99
k = 0.97

# Boolean flag for whether to apply inverse preemphasis
inv_preemphasize = True

ARGS = {
    "wav": wav,
    "k": k,
    "inv_preemphasize": inv_preemphasize,
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
numpy, scipy, torch — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type ndarray, ndim 1, dtype family float, shape (6,) — shape (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
preemphasis Ran
ran on a synthesized input
Rudrabha/Lip2Wav/synthesizer/audio.py
code served (permissive licence) · get_code("56a333de6e89087a")
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 1, dtype family float, shape (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 numpy as np
from scipy import signal

# Create a small audio waveform (wav parameter)
wav = np.array([0.1, 0.2, 0.3, 0.4, 0.5, 0.6], dtype=np.float32)

# Create preemphasis coefficient (k parameter)
k = 0.97

# Create preemphasize flag (boolean parameter)
preemphasize = True

ARGS = {
    "wav": wav,
    "k": k,
    "preemphasize": preemphasize,
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
numpy, scipy — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type ndarray, ndim 1, dtype family float, shape (6,) — shape (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
transform Ran
ran on a synthesized input
Rudrabha/Lip2Wav/face_detection/utils.py
code served (permissive licence) · get_code("fcc1fdc4b492b421")
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 1, dtype family int, shape (2,) 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)

# Create a 2D point as a torch tensor
point = torch.tensor([10.5, 20.3])

# Create center as a numpy array (function accepts both torch.tensor and numpy.array)
center = np.array([50.0, 60.0])

# Define scale as a float
scale = 1.5

# Define resolution as a float
resolution = 64.0

# Define invert as a boolean
invert = False

ARGS = {
    "point": point,
    "center": center,
    "scale": scale,
    "resolution": resolution,
    "invert": invert
}
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 Tensor, ndim 1, dtype family int, shape (2,) — shape (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
draw_gaussian Not yet run
unverified (harvested, no recorded run)
Rudrabha/Lip2Wav/face_detection/utils.py
code served (permissive licence) · get_code("428bab6c5fe3195e")
How it ran

no invocation recorded.

Also present in other papers' code

Of 9 samples lifted from this paper's linked repositories, 9 carry a function identity and 0 do not. Of the 8 distinct functions with an identity, 3 are also present in other papers whose code we hold (3 with at least one paper dated earlier, 0 the same month, 3 later) and 5 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
conv3x3 7 other papers
2 dated earlier, 5 later
1703.07332 (2017-03)
How far are we from solving the 2D & 3D Face Alignment problem? (and a dataset of 230,000 3D facial landmarks)
1807.07860 (2018-07)
Talking Face Generation by Adversarially Disentangled Audio-Visual Representation
2204.02445 (2022-04)
CHORE: Contact, Human and Object REconstruction from a single RGB image
and 4 more not named here
inv_preemphasis 2 other papers
1 dated earlier, 1 later
1803.09017 (2018-03)
Style Tokens: Unsupervised Style Modeling, Control and Transfer in End-to-End Speech Synthesis
2211.14758 (2022-11)
VideoReTalking: Audio-based Lip Synchronization for Talking Head Video Editing In the Wild
preemphasis 2 other papers
1 dated earlier, 1 later
1803.09017 (2018-03)
Style Tokens: Unsupervised Style Modeling, Control and Transfer in End-to-End Speech Synthesis
2211.14758 (2022-11)
VideoReTalking: Audio-based Lip Synchronization for Talking Head Video Editing In the Wild

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 2026-11-21.

Repositories linked to this paper

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

Abstract

Humans involuntarily tend to infer parts of the conversation from lip movements when the speech is absent or corrupted by external noise. In this work, we explore the task of lip to speech synthesis, i.e., learning to generate natural speech given only the lip movements of a speaker. Acknowledging the importance of contextual and speaker-specific cues for accurate lip-reading, we take a different path from existing works. We focus on learning accurate lip sequences to speech mappings for individual speakers in unconstrained, large vocabulary settings. To this end, we collect and release a large-scale benchmark dataset, the first of its kind, specifically to train and evaluate the single-speaker lip to speech task in natural settings. We propose a novel approach with key design choices to achieve accurate, natural lip to speech synthesis in such unconstrained scenarios for the first time. Extensive evaluation using quantitative, qualitative metrics and human evaluation shows that our method is four times more intelligible than previous works in this space. Please check out our demo video for a quick overview of the paper, method, and qualitative results. https://www.youtube.com/watch?v=HziA-jmlk_4&feature=youtu.be

For agents

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

get_harvested_code_for_paper("2005.08209")
get_code_for_paper("2005.08209")
have("2005.08209")

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

curl https://syntology.ai/api/ran/2005.08209.json

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

[![Syntology run record](https://syntology.ai/api/ran/2005.08209.svg)](https://syntology.ai/paper/2005.08209)

Connect an agent — have() is free.