SYNTOLOGY HomeExplorerAtlasCodeMethodologyAboutDevelopersFeedPricing
Paper · 2304.09116 · 2023

NaturalSpeech 2: Latent Diffusion Models are Natural and Zero-Shot Speech and Singing Synthesizers

arXiv · PDF · Open in the Atlas

Code that ran

11 of 11 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, 5 violated it and 5 have no contract verdict; 1 failure 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/2304.09116.json · badge: /api/ran/2304.09116.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
lucidrains/naturalspeech2-pytorch community (classifier label)
classifier basis: plan_row
10 ran · 0 no recorded run
copy not recorded community
not in the paper's declared repositories
1 ran · 0 no recorded run
FunctionStatusWhere it lives
average_over_durations Ran
ran on a synthesized input
lucidrains/naturalspeech2-pytorch/naturalspeech2_pytorch/utils/utils.py
code served (permissive licence) · get_code("040fb4189642c954")
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 (2, 1, 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
from einops import repeat, rearrange

torch.manual_seed(0)

# values: B, 1, T_de (batch, formants, time_decoder)
# durs: B, T_en (batch, time_encoder)
# The function averages values over durations

# Create small tensors with distinct dimensions
batch_size = 2
n_formants = 1
T_de = 7  # decoder time dimension
T_en = 3  # encoder time dimension

# values should be B, 1, T_de with some non-zero values
values = torch.randn(batch_size, n_formants, T_de)

# durs should be B, T_en with positive integers that sum to around T_de
# Each duration specifies how many decoder timesteps correspond to each encoder timestep
durs = torch.tensor([
    [2, 3, 2],  # batch 0: 2+3+2=7
    [2, 2, 3]   # batch 1: 2+2+3=7
], dtype=torch.long)

ARGS = {
    "values": values,
    "durs": durs
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
einops, torch — 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 (2, 1, 3) — shape (2, 1, 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
create_mask Ran
ran on a synthesized input
lucidrains/naturalspeech2-pytorch/naturalspeech2_pytorch/utils/utils.py
code served (permissive licence) · get_code("ac1731a11eef3564")
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, 7) 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
from einops import repeat, rearrange

torch.manual_seed(0)

# sequence_length should be a 1D tensor of shape (batch_size,) containing sequence lengths
# max_len should be an integer representing the maximum sequence length
sequence_length = torch.tensor([3, 5], dtype=torch.long)  # batch_size=2, values represent actual sequence lengths
max_len = 7  # maximum length to create mask up to

ARGS = {
    "sequence_length": sequence_length,
    "max_len": max_len,
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
einops, 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, 7) — shape (2, 7) 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
default Ran
ran, violated the contract we drafted
this paper's copy was not recorded; identical code first harvested from ThomasMrY/VCT
pointer only; the stored copy get_code serves, from ThomasMrY/VCT, is pointer only (licence: NONE) · get_code("60fff7c3c400d7ff")
How it ran

The body of this sample is withheld: this paper's copy of the code was not recorded (its link names no repository or source path), so no licence was read for it. 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.

default Ran
ran, violated the contract we drafted
lucidrains/naturalspeech2-pytorch/naturalspeech2_pytorch/naturalspeech2_pytorch.py
code served (permissive licence) · get_code("d1ef6b8cb9a28a53")
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 str. 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

torch.manual_seed(0)

def exists(x):
    return x is not None

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

# Test case 1: val is not None, should return val
val_1 = "test_value"
d_1 = lambda: "default_value"

# Test case 2: val is None, d is callable, should return d()
val_2 = None
d_2 = lambda: "computed_default"

# Test case 3: val is None, d is not callable, should return d
val_3 = None
d_3 = "static_default"

# Test case 4: val is a number
val_4 = 42
d_4 = 100

# Test case 5: val is a tensor
val_5 = torch.tensor([1.0, 2.0, 3.0])
d_5 = lambda: torch.tensor([0.0])

ARGS = {
    "val": val_1,
    "d": d_1
}
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 str — a str output has no array shape of its own; none is stated — the instrument coerces any output to an array; for a str 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
divisible_by Ran
ran, violated the contract we drafted
lucidrains/naturalspeech2-pytorch/naturalspeech2_pytorch/naturalspeech2_pytorch.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
lucidrains/naturalspeech2-pytorch/naturalspeech2_pytorch/aligner.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.

exists Ran
ran, violated the contract we drafted
lucidrains/naturalspeech2-pytorch/naturalspeech2_pytorch/naturalspeech2_pytorch.py
code served (permissive licence) · get_code("608e364a9d2376a3")
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

def exists(x):
    return x is not None

ARGS = {
    "x": torch.tensor([1, 2, 3, 4])
}
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 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
is_tool Ran
ran on a synthesized input
lucidrains/naturalspeech2-pytorch/naturalspeech2_pytorch/utils/phonemizers/espeak_wrapper.py
code served (permissive licence) · get_code("a57ad0dfe43eabea")
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
import numpy as np

def is_tool(name):
    from shutil import which
    return which(name) is not None

# The function takes a single parameter 'name' which should be a string
# representing the name of a tool/command to check for in the system PATH
ARGS = {
    "name": "python"
}
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 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
maximum_path Ran
ran on a synthesized input
lucidrains/naturalspeech2-pytorch/naturalspeech2_pytorch/aligner.py
code served (permissive licence) · get_code("bfae004145c18b24")
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 (2, 4, 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 torch.nn.functional as F

torch.manual_seed(0)

# Create small test tensors for maximum_path function
batch_size = 2
t_x = 4
t_y = 3

# value: shape (b, t_x, t_y) - the values to compute maximum path over
value = torch.randn(batch_size, t_x, t_y, dtype=torch.float32)

# mask: shape (b, t_x, t_y) - binary mask indicating valid positions
mask = torch.ones(batch_size, t_x, t_y, dtype=torch.float32)

# const: optional constant for padding (default None, will be set to -inf inside function)
const = None

ARGS = {
    "value": value,
    "mask": mask,
    "const": const
}
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 3, dtype family float, shape (2, 4, 3) — shape (2, 4, 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
once Ran
ran; our contract draft was wrong, not the code
lucidrains/naturalspeech2-pytorch/naturalspeech2_pytorch/attend.py
code served (permissive licence) · get_code("52e10a7d6b0fa036")
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 function. 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 functools import wraps

def once(fn):
    called = False

    @wraps(fn)
    def inner(x):
        nonlocal called
        if called:
            return
        called = True
        return fn(x)
    return inner


def sample_function(x):
    """A simple function that the decorator will wrap."""
    return x * 2


ARGS = {
    "fn": sample_function
}
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 function — a function output has no array shape of its own; none is stated — the instrument coerces any output to an array; for a function 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
pad_tensor Ran
ran on a synthesized input
lucidrains/naturalspeech2-pytorch/naturalspeech2_pytorch/aligner.py
code served (permissive licence) · get_code("c6289707dcc4058f")
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 (4, 7, 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.nn.functional as F

torch.manual_seed(0)

# Create a small input tensor
# Using shape (2, 3, 4) for a 3D tensor to test padding
input_tensor = torch.randn(2, 3, 4)

# Create padding specification
# pad is a list of tuples, one for each dimension (in reverse order)
# For a 3D tensor, we need 3 tuples: (pad_left, pad_right) for each dim
# Reverse order means: last dim first, then middle dim, then first dim
pad = [(1, 1), (2, 2), (0, 1)]  # pad last dim by 1 on each side, middle dim by 2, first dim by 0,1

# Padding value
value = 0

ARGS = {
    'input': input_tensor,
    'pad': pad,
    'value': value
}
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 3, dtype family float, shape (4, 7, 5) — shape (4, 7, 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

Also present in other papers' code

Of 11 samples lifted from this paper's linked repositories, 11 carry a function identity and 0 do not. Of the 11 distinct functions with an identity, 6 are also present in other papers whose code we hold (6 with at least one paper dated earlier, 3 the same month, 6 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
exists 486 other papers
166 dated earlier, 7 the same month, 307 later, 4 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 163 other papers
80 dated earlier, 1 the same month, 80 later, 1 same year, order not determinable, 1 undated · 27 only as an assembly brick
1603.08155 (2016-03)
Perceptual Losses for Real-Time Style Transfer and Super-Resolution
1810.00825 (2018-10)
Set Transformer: A Framework for Attention-based Permutation-Invariant Neural Networks
1907.01470 (2019-07)
Augmenting Self-attention with Persistent Memory
and 160 more not named here
exists 144 other papers
35 dated earlier, 1 the same month, 107 later, 1 same year, order not determinable · 40 only as an assembly brick
1611.07004 (2016-11)
Image-to-Image Translation with Conditional Adversarial Networks
1711.11585 (2017-11)
High-Resolution Image Synthesis and Semantic Manipulation with Conditional GANs
2006.11239 (2020-06)
Denoising Diffusion Probabilistic Models
and 141 more not named here
default 79 other papers
16 dated earlier, 63 later · 15 only as an assembly brick
1611.07004 (2016-11)
Image-to-Image Translation with Conditional Adversarial Networks
2006.11239 (2020-06)
Denoising Diffusion Probabilistic Models
2010.02502 (2020-10)
Denoising Diffusion Implicit Models
and 76 more not named here
once 19 other papers
8 dated earlier, 11 later · 4 only as an assembly brick
2006.11239 (2020-06)
Denoising Diffusion Probabilistic Models
2202.01855 (2022-02)
Self-Supervised Learning with Random-Projection Quantizer for Speech Recognition
2203.07852 (2022-03)
Block-Recurrent Transformers
and 16 more not named here
divisible_by 16 other papers
3 dated earlier, 13 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

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

Scaling text-to-speech (TTS) to large-scale, multi-speaker, and in-the-wild datasets is important to capture the diversity in human speech such as speaker identities, prosodies, and styles (e.g., singing). Current large TTS systems usually quantize speech into discrete tokens and use language models to generate these tokens one by one, which suffer from unstable prosody, word skipping/repeating issue, and poor voice quality. In this paper, we develop NaturalSpeech 2, a TTS system that leverages a neural audio codec with residual vector quantizers to get the quantized latent vectors and uses a diffusion model to generate these latent vectors conditioned on text input. To enhance the zero-shot capability that is important to achieve diverse speech synthesis, we design a speech prompting mechanism to facilitate in-context learning in the diffusion model and the duration/pitch predictor. We scale NaturalSpeech 2 to large-scale datasets with 44K hours of speech and singing data and evaluate its voice quality on unseen speakers. NaturalSpeech 2 outperforms previous TTS systems by a large margin in terms of prosody/timbre similarity, robustness, and voice quality in a zero-shot setting, and performs novel zero-shot singing synthesis with only a speech prompt. Audio samples are available at https://speechresearch.github.io/naturalspeech2.

For agents

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

get_harvested_code_for_paper("2304.09116")
get_code_for_paper("2304.09116")
have("2304.09116")

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

curl https://syntology.ai/api/ran/2304.09116.json

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

[![Syntology run record](https://syntology.ai/api/ran/2304.09116.svg)](https://syntology.ai/paper/2304.09116)

Connect an agent — have() is free.