SYNTOLOGY HomeExplorerAtlasCodeMethodologyAboutDevelopersFeedPricing
Paper · 2307.09323 · ICCV · 2023

Efficient Region-Aware Neural Radiance Fields for High-Fidelity Talking Portrait Synthesis

Jiawei Zhang, Jun Zhou, Lin Gu, Jiahe Li, Xiao Bai

arXiv · PDF · Open in the Atlas

Code that ran

10 of 12 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; 10 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/2307.09323.json · badge: /api/ran/2307.09323.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
fictionarry/er-nerf canonical (classifier label)
classifier basis: census
10 ran · 2 no recorded run
FunctionStatusWhere it lives
conv3x3 Ran
ran; our contract draft was wrong, not the code
fictionarry/er-nerf/data_utils/face_parsing/resnet.py
pointer only (licence: MIT); the stored copy get_code serves, from lixiaotong97/DSU, is pointer only (licence: NONE) · get_code("fac5364e2f53c6db")
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 lixiaotong97/DSU, 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.

create_layer_basic Ran
ran; our contract draft was wrong, not the code
fictionarry/er-nerf/data_utils/face_parsing/resnet.py
pointer only (licence: MIT); the stored copy get_code serves, from vladhondru25/DF-MoE, is pointer only (licence: NOASSERTION) · get_code("325b8d54f52e2421")
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 vladhondru25/DF-MoE, and GitHub could not classify the licence of that repository (NOASSERTION). 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_audio_features Ran
ran; the contract check could not run (RAISES, DEP_MISSING or TIMEOUT)
fictionarry/er-nerf/nerf_triplane/utils.py
code served (permissive licence) · get_code("083a6ac840310b58")
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, 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)

# Create a small audio features tensor
# Shape: (sequence_length, feature_dim)
features = torch.randn(20, 4)

# Test with att_mode = 0, 1, or 2
# We'll use att_mode = 1 and index = 10 as an example
att_mode = 1
index = 10

ARGS = {
    'features': features,
    'att_mode': att_mode,
    'index': index
}
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, 4) — shape (8, 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
linear_to_srgb Ran
ran; the contract check could not run (RAISES, DEP_MISSING or TIMEOUT)
fictionarry/er-nerf/nerf_triplane/utils.py
code served (permissive licence) · get_code("df2164d2c1f74c1b")
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; 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)

# The function takes a single parameter 'x' which is a tensor
# It applies element-wise operations: comparison, multiplication, exponentiation
# So x should be a tensor with numeric values in a reasonable range for color space conversion

x = torch.tensor([
    [0.0, 0.001, 0.0031308, 0.5],
    [0.2, 0.8, 1.0, 0.1]
], dtype=torch.float32)

ARGS = {
    "x": x
}
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 — no output shape is stated: observe_signature.py could not observe it (import:OSError)
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
make_even_first_dim Ran
ran; our contract draft was wrong, not the code
fictionarry/er-nerf/data_utils/hubert.py
pointer only (licence: MIT); the stored copy get_code serves, from ZiqiaoPeng/SyncTalk, is pointer only (licence: NOASSERTION) · get_code("5674b30130f49b38")
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 ZiqiaoPeng/SyncTalk, and GitHub could not classify the licence of that repository (NOASSERTION). 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.

nerf_matrix_to_ngp Ran
ran; the contract check could not run (RAISES, DEP_MISSING or TIMEOUT)
fictionarry/er-nerf/nerf_triplane/provider.py
code served (permissive licence) · get_code("cc5a245ddce1b228")
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 (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 numpy as np

# Create a sample pose matrix (4x4 transformation matrix)
# The function expects to index rows 0, 1, 2 and access columns 0, 1, 2, 3
pose = np.array([
    [1.0, 0.0, 0.0, 5.0],
    [0.0, 1.0, 0.0, 6.0],
    [0.0, 0.0, 1.0, 7.0],
    [0.0, 0.0, 0.0, 1.0]
], dtype=np.float32)

ARGS = {
    "pose": pose,
    "scale": 0.33,
    "offset": [0, 0, 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 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
polygon_area Ran
ran; the contract check could not run (RAISES, DEP_MISSING or TIMEOUT)
fictionarry/er-nerf/nerf_triplane/provider.py
code served (permissive licence) · get_code("5e3067f7bf8c9e33")
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 float64, ndim 0, 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 numpy as np

# The function takes x and y coordinates of polygon vertices
# It computes the area using the shoelace formula
# x and y should be 1D numpy arrays of the same length

x = np.array([0.0, 1.0, 1.0, 0.0])
y = np.array([0.0, 0.0, 1.0, 1.0])

ARGS = {
    "x": x,
    "y": y,
}
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 float64, ndim 0, dtype family float — no output shape is stated: observe_signature.py recorded float ndim 0, which disagrees with the fixture run
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
sample_pdf Ran
ran; our contract draft was wrong, not the code
fictionarry/er-nerf/nerf_triplane/renderer.py
code served (permissive licence) · get_code("8ed7b9fc7f33d1dd")
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 (2, 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

torch.manual_seed(0)

# Construct arguments for sample_pdf
# The function expects:
# - bins: tensor of bin edges
# - weights: tensor of weights for each bin
# - n_samples: number of samples to draw
# - det: boolean for deterministic sampling

# Create small tensors with distinct dimensions
# bins: shape (batch=2, num_bins=5) - the bin edges
bins = torch.linspace(0, 1, steps=5).unsqueeze(0).expand(2, -1)

# weights: shape (batch=2, num_bins=4) - weights for each bin
# Note: num_bins for weights should be one less than bins since weights are for intervals
weights = torch.rand(2, 4)

# n_samples: number of samples to draw
n_samples = 3

# det: deterministic flag
det = False

ARGS = {
    "bins": bins,
    "weights": weights,
    "n_samples": n_samples,
    "det": det,
}
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 (2, 3) — shape (2, 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
smooth_camera_path Ran
ran; the contract check could not run (RAISES, DEP_MISSING or TIMEOUT)
fictionarry/er-nerf/nerf_triplane/provider.py
code served (permissive licence) · get_code("c08d25b3e8575c34")
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 (3, 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 numpy as np
from scipy.spatial.transform import Slerp, Rotation

# Create a small batch of camera poses
# poses should be shape (N, 4, 4) representing N camera transformation matrices
# Each pose is a 4x4 homogeneous transformation matrix
N = 3  # Number of camera poses
poses = np.zeros((N, 4, 4), dtype=np.float32)

# Initialize as identity matrices with some variation
for i in range(N):
    poses[i] = np.eye(4)
    # Add some translation variation
    poses[i, :3, 3] = np.array([i * 0.1, i * 0.2, i * 0.3], dtype=np.float32)
    # Add small rotation variation
    angle = i * 0.1
    rot_matrix = np.array([
        [np.cos(angle), -np.sin(angle), 0],
        [np.sin(angle), np.cos(angle), 0],
        [0, 0, 1]
    ], dtype=np.float32)
    poses[i, :3, :3] = rot_matrix

ARGS = {
    "poses": poses,
    "kernel_size": 3
}
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 3, dtype family float, shape (3, 4, 4) — shape (3, 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
srgb_to_linear Ran
ran; the contract check could not run (RAISES, DEP_MISSING or TIMEOUT)
fictionarry/er-nerf/nerf_triplane/utils.py
code served (permissive licence) · get_code("78534c794e6d4ae8")
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; 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)

# The function takes a tensor x with sRGB values and converts to linear
# It uses torch.where to conditionally apply two formulas based on threshold
# No other parameters or attributes are accessed

x = torch.rand(2, 4)  # Small tensor with values in [0, 1]

ARGS = {
    "x": x
}
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 — no output shape is stated: observe_signature.py could not observe it (import:OSError)
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_encoder Not yet run
unverified (harvested, no recorded run)
fictionarry/er-nerf/encoding.py
pointer only (licence: MIT); the stored copy get_code serves, from ZiqiaoPeng/SyncTalk, is pointer only (licence: NOASSERTION) · get_code("c772127017598c32")
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 ZiqiaoPeng/SyncTalk, and GitHub could not classify the licence of that repository (NOASSERTION). 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. No invocation recorded.

save_transforms Not yet run
unverified (harvested, no recorded run)
fictionarry/er-nerf/data_utils/process.py
code served (permissive licence) · get_code("874fee02f29df197")
How it ran

no invocation recorded.

Also present in other papers' code

Of 12 samples lifted from this paper's linked repositories, 12 carry a function identity and 0 do not. Of the 12 distinct functions with an identity, 12 are also present in other papers whose code we hold (11 with at least one paper dated earlier, 1 the same month, 5 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
conv3x3 686 other papers
489 dated earlier, 8 the same month, 182 later, 7 same year, order not determinable · 31 only as an assembly brick
1406.2661 (2014-06)
Generative Adversarial Networks
1409.0575 (2014-09)
ImageNet Large Scale Visual Recognition Challenge
1412.6572 (2014-12)
Explaining and Harnessing Adversarial Examples
and 683 more not named here
create_layer_basic 18 other papers
10 dated earlier, 8 later
1904.07399 (2019-04)
Adaptive Wing Loss for Robust Face Alignment via Heatmap Regression
1911.09070 (2019-11)
EfficientDet: Scalable and Efficient Object Detection
2011.00844 (2020-11)
Do 2D GANs Know 3D Shape? Unsupervised 3D shape reconstruction from 2D Image GANs
and 15 more not named here
sample_pdf 10 other papers
8 dated earlier, 1 later, 1 same year, order not determinable · 1 only as an assembly brick
2110.00168 (2021-10)
Vision-Only Robot Navigation in a Neural Radiance World
2112.01455 (2021-12)
Zero-Shot Text-Guided Object Generation with Dream Fields
2203.09517 (2022-03)
TensoRF: Tensorial Radiance Fields
and 7 more not named here
linear_to_srgb 4 other papers
3 dated earlier, 1 same year, order not determinable
2110.00168 (2021-10)
Vision-Only Robot Navigation in a Neural Radiance World
2211.12368 (2022-11)
Real-time Neural Radiance Talking Portrait Synthesis via Audio-spatial Decomposition
aaai_25135 (2023 (year only))
One Is All: Bridging the Gap between Neural Radiance Fields Architectures with Progressive Volume Distillation
and 1 more not named here
srgb_to_linear 4 other papers
3 dated earlier, 1 same year, order not determinable
2110.00168 (2021-10)
Vision-Only Robot Navigation in a Neural Radiance World
2211.12368 (2022-11)
Real-time Neural Radiance Talking Portrait Synthesis via Audio-spatial Decomposition
aaai_25135 (2023 (year only))
One Is All: Bridging the Gap between Neural Radiance Fields Architectures with Progressive Volume Distillation
and 1 more not named here
get_encoder 2 other papers
1 dated earlier, 1 later
2211.12368 (2022-11)
Real-time Neural Radiance Talking Portrait Synthesis via Audio-spatial Decomposition
2311.17590 (2023-11)
SyncTalk: The Devil is in the Synchronization for Talking Head Synthesis
nerf_matrix_to_ngp 2 other papers
2 dated earlier
2110.00168 (2021-10)
Vision-Only Robot Navigation in a Neural Radiance World
2211.12368 (2022-11)
Real-time Neural Radiance Talking Portrait Synthesis via Audio-spatial Decomposition
get_audio_features 1 other paper
1 dated earlier
2211.12368 (2022-11)
Real-time Neural Radiance Talking Portrait Synthesis via Audio-spatial Decomposition
make_even_first_dim 1 other paper
1 later
2311.17590 (2023-11)
SyncTalk: The Devil is in the Synchronization for Talking Head Synthesis
polygon_area 1 other paper
1 dated earlier
2211.12368 (2022-11)
Real-time Neural Radiance Talking Portrait Synthesis via Audio-spatial Decomposition
save_transforms 1 other paper
1 dated earlier
2211.12368 (2022-11)
Real-time Neural Radiance Talking Portrait Synthesis via Audio-spatial Decomposition
smooth_camera_path 1 other paper
1 dated earlier
2211.12368 (2022-11)
Real-time Neural Radiance Talking Portrait Synthesis via Audio-spatial Decomposition

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

This paper presents ER-NeRF, a novel conditional Neural Radiance Fields (NeRF) based architecture for talking portrait synthesis that can concurrently achieve fast convergence, real-time rendering, and state-of-the-art performance with small model size. Our idea is to explicitly exploit the unequal contribution of spatial regions to guide talking portrait modeling. Specifically, to improve the accuracy of dynamic head reconstruction, a compact and expressive NeRF-based Tri-Plane Hash Representation is introduced by pruning empty spatial regions with three planar hash encoders. For speech audio, we propose a Region Attention Module to generate region-aware condition feature via an attention mechanism. Different from existing methods that utilize an MLP-based encoder to learn the cross-modal relation implicitly, the attention mechanism builds an explicit connection between audio features and spatial regions to capture the priors of local motions. Moreover, a direct and fast Adaptive Pose Encoding is introduced to optimize the head-torso separation problem by mapping the complex transformation of the head pose into spatial coordinates. Extensive experiments demonstrate that our method renders better high-fidelity and audio-lips synchronized talking portrait videos, with realistic details and high efficiency compared to previous methods. Code is available at https://github.com/Fictionarry/ ER-NeRF.

For agents

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

get_harvested_code_for_paper("2307.09323")
get_code_for_paper("2307.09323")
have("2307.09323")

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

curl https://syntology.ai/api/ran/2307.09323.json

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

[![Syntology run record](https://syntology.ai/api/ran/2307.09323.svg)](https://syntology.ai/paper/2307.09323)

Connect an agent — have() is free.