SYNTOLOGY HomeExplorerAtlasCodeMethodologyAboutDevelopersFeedPricing
Paper · 2009.01776 · 2020

HiFiSinger: Towards High-Fidelity Neural Singing Voice Synthesis

arXiv · PDF · Open in the Atlas

Code that ran

7 of 10 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; 0 failures of our instrument; 3 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/2009.01776.json · badge: /api/ran/2009.01776.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
CODEJIN/HiFiSinger community (classifier label)
classifier basis not recorded on the link
7 ran · 3 no recorded run
FunctionStatusWhere it lives
Mel_Stack Ran
ran on a synthesized input
CODEJIN/HiFiSinger/Datasets.py
code served (permissive licence) · get_code("9f069d3d7349b7e4")
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, 7, 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 sample mel-spectrograms with different lengths
mel1 = np.random.randn(5, 4)  # 5 time steps, 4 frequency bins
mel2 = np.random.randn(3, 4)  # 3 time steps, 4 frequency bins
mel3 = np.random.randn(7, 4)  # 7 time steps, 4 frequency bins

mels = [mel1, mel2, mel3]
max_abs_mel = 2.5

ARGS = {
    "mels": mels,
    "max_abs_mel": max_abs_mel,
}
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 3, dtype family float, shape (3, 7, 4) — shape (3, 7, 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
Preemphasis Ran
ran on a synthesized input
CODEJIN/HiFiSinger/Audio.py
code served (permissive licence) · get_code("494d58dba436fa48")
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

torch.manual_seed(0)

# The function takes audio and pre_emphasis as parameters
# audio should be array-like (numpy array or list) that can be passed to signal.lfilter
# pre_emphasis is a float with default 0.97

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

# Define the arguments dictionary
ARGS = {
    "audio": audio,
    "pre_emphasis": 0.97
}
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
Text_to_Token Ran
ran on a synthesized input
CODEJIN/HiFiSinger/Datasets.py
code served (permissive licence) · get_code("436aa29055fdc19e")
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 int; 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 list of text items and a dictionary mapping text to tokens
# It returns a list where each text item is replaced by its token from the dictionary

text = ['hello', 'world', 'hello']
token_dict = {
    'hello': 1,
    'world': 2
}

ARGS = {
    'text': text,
    'token_dict': token_dict
}
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 list, ndim 1, dtype family int — 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
cumulativeMeanNormalizedDifferenceFunction Ran
ran on a synthesized input
CODEJIN/HiFiSinger/yin.py
code served (permissive licence) · get_code("e507922638a3a56e")
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

# Create a small difference function array
# The function expects df to be an array-like object with numerical values
# N represents the length of the original data
df = np.array([1.0, 0.8, 0.6, 0.5, 0.4, 0.3])
N = len(df)

ARGS = {
    "df": df,
    "N": N
}
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 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
differenceFunction Ran
ran on a synthesized input
CODEJIN/HiFiSinger/yin.py
code served (permissive licence) · get_code("3cbbf9560f72cc60")
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 (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

# Construct arguments for differenceFunction
# The function expects:
# - x: audio data (array-like)
# - N: length of data (int)
# - tau_max: integration window size (int)

# Create small audio data
audio_data = np.array([0.1, 0.2, 0.15, 0.3, 0.25, 0.1, 0.05, 0.2], dtype=np.float64)
N = len(audio_data)
tau_max = 4

ARGS = {
    "x": audio_data,
    "N": N,
    "tau_max": tau_max
}
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 1, dtype family float, shape (4,) — shape (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
getPitch Ran
ran on a synthesized input
CODEJIN/HiFiSinger/yin.py
code served (permissive licence) · get_code("762aa23bfe00152c")
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 int, ndim 0, dtype family int, 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 expects:
# - cmdf: array-like object that supports indexing with integers
# - tau_min: integer (minimum period)
# - tau_max: integer (maximum period)
# - harmo_th: float (threshold, default 0.1)

# Create a small CMDF array (Cumulative Mean Normalized Difference function)
# It needs to be indexable and contain float values
# We'll create an array with some values below and above the threshold
cmdf_values = np.array([
    0.5,   # tau=0
    0.15,  # tau=1 (below threshold of 0.1? No, above)
    0.08,  # tau=2 (below threshold)
    0.07,  # tau=3 (below threshold, and below previous)
    0.09,  # tau=4 (above previous, so loop would stop)
    0.2,   # tau=5
    0.3,   # tau=6
    0.4,   # tau=7
    0.5,   # tau=8
], dtype=np.float32)

tau_min = 1
tau_max = 8
harmo_th = 0.1

ARGS = {
    "cmdf": cmdf_values,
    "tau_min": tau_min,
    "tau_max": tau_max,
    "harmo_th": harmo_th
}
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 int, ndim 0, dtype family int, 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
reduce_tensor Ran
ran on a synthesized input
CODEJIN/HiFiSinger/distributed.py
code served (permissive licence) · get_code("a9bfcbb1d3623ba1")
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, 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
import torch.distributed as dist
from unittest.mock import MagicMock, patch

torch.manual_seed(0)

# Mock dist.all_reduce and dist.ReduceOp since we're in a CPU-only environment
# without actual distributed setup
original_all_reduce = dist.all_reduce
original_reduce_op = dist.ReduceOp

def mock_all_reduce(tensor, op=None):
    """Mock all_reduce that just returns without doing distributed ops"""
    pass

# Patch the distributed module for this execution
dist.all_reduce = mock_all_reduce

# Create a small tensor to pass
tensor = torch.randn(2, 4)

# Number of GPUs
num_gpus = 2

ARGS = {
    "tensor": tensor,
    "num_gpus": num_gpus,
}
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, 4) — shape (2, 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
Recursive_Parse Not yet run
unverified (harvested, no recorded run)
CODEJIN/HiFiSinger/Arg_Parser.py
code served (permissive licence) · get_code("564512b44009385c")
How it ran

no invocation recorded.

Silence_Stack Not yet run
unverified (harvested, no recorded run)
CODEJIN/HiFiSinger/Datasets.py
code served (permissive licence) · get_code("a944e54ed971b982")
How it ran

no invocation recorded.

apply_gradient_allreduce Not yet run
unverified (harvested, no recorded run)
CODEJIN/HiFiSinger/distributed.py
code served (permissive licence) · get_code("360725578b4657a7")
How it ran

no invocation recorded.

Also present in other papers' code

Of 10 samples lifted from this paper's linked repositories, 10 carry a function identity and 0 do not. Of the 10 distinct functions with an identity, 6 are also present in other papers whose code we hold (6 with at least one paper dated earlier, 0 the same month, 0 later) and 4 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
apply_gradient_allreduce 2 other papers
2 dated earlier
1811.00002 (2018-11)
WaveGlow: A Flow-based Generative Network for Speech Synthesis
2005.05957 (2020-05)
Flowtron: an Autoregressive Flow-based Generative Network for Text-to-Speech Synthesis
Preemphasis 1 other paper
1 dated earlier
2004.11284 (2020-04)
Unsupervised Speech Decomposition via Triple Information Bottleneck
cumulativeMeanNormalizedDifferenceFunction 1 other paper
1 dated earlier
2004.11284 (2020-04)
Unsupervised Speech Decomposition via Triple Information Bottleneck
differenceFunction 1 other paper
1 dated earlier
2004.11284 (2020-04)
Unsupervised Speech Decomposition via Triple Information Bottleneck
getPitch 1 other paper
1 dated earlier
2004.11284 (2020-04)
Unsupervised Speech Decomposition via Triple Information Bottleneck
reduce_tensor 1 other paper
1 dated earlier
2005.05957 (2020-05)
Flowtron: an Autoregressive Flow-based Generative Network for Text-to-Speech Synthesis

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

High-fidelity singing voices usually require higher sampling rate (e.g., 48kHz) to convey expression and emotion. However, higher sampling rate causes the wider frequency band and longer waveform sequences and throws challenges for singing voice synthesis (SVS) in both frequency and time domains. Conventional SVS systems that adopt small sampling rate cannot well address the above challenges. In this paper, we develop HiFiSinger, an SVS system towards high-fidelity singing voice. HiFiSinger consists of a FastSpeech based acoustic model and a Parallel WaveGAN based vocoder to ensure fast training and inference and also high voice quality. To tackle the difficulty of singing modeling caused by high sampling rate (wider frequency band and longer waveform), we introduce multi-scale adversarial training in both the acoustic model and vocoder to improve singing modeling. Specifically, 1) To handle the larger range of frequencies caused by higher sampling rate, we propose a novel sub-frequency GAN (SF-GAN) on mel-spectrogram generation, which splits the full 80-dimensional mel-frequency into multiple sub-bands and models each sub-band with a separate discriminator. 2) To model longer waveform sequences caused by higher sampling rate, we propose a multi-length GAN (ML-GAN) for waveform generation to model different lengths of waveform sequences with separate discriminators. 3) We also introduce several additional designs and findings in HiFiSinger that are crucial for high-fidelity voices, such as adding F0 (pitch) and V/UV (voiced/unvoiced flag) as acoustic features, choosing an appropriate window/hop size for mel-spectrogram, and increasing the receptive field in vocoder for long vowel modeling. Experiment results show that HiFiSinger synthesizes high-fidelity singing voices with much higher quality: 0.32/0.44 MOS gain over 48kHz/24kHz baseline and 0.83 MOS gain over previous SVS systems.

For agents

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

get_harvested_code_for_paper("2009.01776")
get_code_for_paper("2009.01776")
have("2009.01776")

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

curl https://syntology.ai/api/ran/2009.01776.json

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

[![Syntology run record](https://syntology.ai/api/ran/2009.01776.svg)](https://syntology.ai/paper/2009.01776)

Connect an agent — have() is free.