SYNTOLOGY HomeExplorerAtlasCodeMethodologyAboutDevelopersFeedPricing
Paper · 2205.04421 · 2022

NaturalSpeech: End-to-End Text to Speech Synthesis with Human-Level Quality

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 3 have no contract verdict; 0 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/2205.04421.json · badge: /api/ran/2205.04421.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
daniilrobnikov/vits2 community (classifier label)
classifier basis not recorded on the link
3 ran · 0 no recorded run
FunctionStatusWhere it lives
discriminator_loss Ran
ran on a synthesized input
daniilrobnikov/vits2/losses.py
code served (permissive licence) · get_code("c55e3d6390482f28")
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 tuple. 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)

# Create small discriminator outputs for real and generated data
# The function expects lists of tensors that will be converted to float and squeezed into scalars via mean()
disc_real_outputs = [
    torch.randn(2, 4),
    torch.randn(2, 3),
]

disc_generated_outputs = [
    torch.randn(2, 4),
    torch.randn(2, 3),
]

ARGS = {
    "disc_real_outputs": disc_real_outputs,
    "disc_generated_outputs": disc_generated_outputs,
}
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 tuple — a tuple output has no array shape of its own; none is 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
feature_loss Ran
ran on a synthesized input
daniilrobnikov/vits2/losses.py
code served (permissive licence) · get_code("f9a2eedb621a3dda")
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 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
from typing import List

torch.manual_seed(0)

# Create two lists of feature maps
# Each list contains multiple "layers" (dr, dg), and each layer contains multiple "scales" (rl, gl)
# The function iterates: for dr, dg in zip(fmap_r, fmap_g) then for rl, gl in zip(dr, dg)
# So fmap_r and fmap_g are lists of lists of tensors

# Create first feature map list (from real/reference model)
fmap_r = [
    [torch.randn(2, 4), torch.randn(2, 3)],  # layer 1: 2 tensors
    [torch.randn(2, 5), torch.randn(2, 6)]   # layer 2: 2 tensors
]

# Create second feature map list (from generated/predicted model)
fmap_g = [
    [torch.randn(2, 4), torch.randn(2, 3)],  # layer 1: 2 tensors
    [torch.randn(2, 5), torch.randn(2, 6)]   # layer 2: 2 tensors
]

ARGS = {
    "fmap_r": fmap_r,
    "fmap_g": fmap_g
}
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 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
generator_loss Ran
ran on a synthesized input
daniilrobnikov/vits2/losses.py
code served (permissive licence) · get_code("3eac5c1dd8f1b229")
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 tuple. 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 typing import List

torch.manual_seed(0)

# The function expects a list of tensors (discriminator outputs)
# It iterates through them, converts to float, and computes loss
# Each tensor should be a batch of discriminator outputs (logits/probabilities)

disc_output_1 = torch.randn(2, 4)  # Small batch of 2, 4 discriminator outputs
disc_output_2 = torch.randn(2, 3)  # Different dimension for distinction
disc_output_3 = torch.randn(2, 5)  # Another batch with different size

disc_outputs = [disc_output_1, disc_output_2, disc_output_3]

ARGS = {
    "disc_outputs": disc_outputs
}
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 tuple — a tuple output has no array shape of its own; none is 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

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, 0 are also present in other papers whose code we hold (0 with at least one paper dated earlier, 0 the same month, 0 later) and 3 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.

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

Text to speech (TTS) has made rapid progress in both academia and industry in recent years. Some questions naturally arise that whether a TTS system can achieve human-level quality, how to define/judge that quality and how to achieve it. In this paper, we answer these questions by first defining the human-level quality based on the statistical significance of subjective measure and introducing appropriate guidelines to judge it, and then developing a TTS system called NaturalSpeech that achieves human-level quality on a benchmark dataset. Specifically, we leverage a variational autoencoder (VAE) for end-to-end text to waveform generation, with several key modules to enhance the capacity of the prior from text and reduce the complexity of the posterior from speech, including phoneme pre-training, differentiable duration modeling, bidirectional prior/posterior modeling, and a memory mechanism in VAE. Experiment evaluations on popular LJSpeech dataset show that our proposed NaturalSpeech achieves -0.01 CMOS (comparative mean opinion score) to human recordings at the sentence level, with Wilcoxon signed rank test at p-level p >> 0.05, which demonstrates no statistically significant difference from human recordings for the first time on this dataset.

For agents

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

get_harvested_code_for_paper("2205.04421")
get_code_for_paper("2205.04421")
have("2205.04421")

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

curl https://syntology.ai/api/ran/2205.04421.json

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

[![Syntology run record](https://syntology.ai/api/ran/2205.04421.svg)](https://syntology.ai/paper/2205.04421)

Connect an agent — have() is free.