Papers › High-Fidelity Audio Compression with Improved RVQGAN

High-Fidelity Audio Compression with Improved RVQGAN

11 Jun 2023NeurIPS 2023 11arXiv:2306.06546archive 2025-07-28

Rithesh Kumar, Prem Seetharaman, Alejandro Luebs, Ishaan Kumar, Kundan Kumar

Language models have been successfully used to model natural signals, such as images, speech, and music. A key component of these models is a high quality neural compression model that can compress high-dimensional natural signals into lower dimensional discrete tokens. To that end, we introduce a high-fidelity universal neural audio compression algorithm that achieves ~90x compression of 44.1 KHz audio into tokens at just 8kbps bandwidth. We achieve this by combining advances in high-fidelity audio generation with better vector quantization techniques from the image domain, along with improved adversarial and reconstruction losses. We compress all domains (speech, environment, music, etc.) with a single universal model, making it widely applicable to generative modeling of all audio. We compare with competing audio compression algorithms, and find our method outperforms them significantly. We provide thorough ablations for every design choice, as well as open-source code and trained model weights. We hope our work can lay the foundation for the next generation of high-fidelity audio modeling.

PaperPDFConference PDFCodeCode Syntology ranMethod as code

In Syntology View this paper on Syntology, its page in Syntology's graph. That page adds which of its functions are also present in other papers' code Syntology holds, with dates, and its dated run record as JSON and a README badge. It also lists the repositories linked to the paper, the abstract and the calls for agents.

Open this paper in Syntology's Atlas, the map of the papers in Syntology's graph and their citations.

For agents, on Syntology's MCP service (how to connect):

Code

No sample from the archive's official repository (descriptinc/descript-audio-codec) is linked to this paper; the 38 samples below come from 2 community repositories. Syntology harvested code from descriptinc/descript-audio-codec for another paper.

descriptinc/descript-audio-codec: Syntology's graph links these samples to The Equalizer: Introducing Shape-Gain Decomposition in Neural Audio Codecs (arXiv:2602.15491; that paper's own run record: 8 ran (of which 0 constructed an object rather than computing a result; 2 with no instrument failure: 0 honoured, 1 violated, 1 with no contract checked; 6 where Syntology's instrument failed) · 2 unverified); nothing checks that they implement this paper's method.

Syntology 38 code samples harvested from 2 repositories linked to this paper: 27 ran (of which 10 constructed an object rather than computing a result; 17 with no instrument failure: 3 honoured, 3 violated, 11 with no contract checked; 10 where Syntology's instrument failed) · 11 unverified. Where Syntology's instrument failed: 4 where our contract draft was wrong; 6 where the contract check could not run; failures of Syntology's instrument, not of the code.

By repository: community (archive-listed): 38 samples from 2 repositories, 27 ran (17 with no instrument failure; 10 where Syntology's instrument failed). The run record, sample by sample. “Ran” means executed on a synthesized input, not that the code is correct or reproduces the paper.

No sample is linked to this paper from 2 of its repositories that Syntology harvested for other papers:

Levels written between 2026-09-22 and 2026-09-24 (UTC): the dates Syntology's graph wrote each sample's verification level, not when the code executed.

descriptinc/descript-audio-codec officialmentioned in paperpytorchMITSyntology: no sample linked to this paper (harvested for another paper). Syntology's graph links these samples to The Equalizer: Introducing Shape-Gain Decomposition in Neural Audio Codecs (arXiv:2602.15491; that paper's own run record: 8 ran (of which 0 constructed an object rather than computing a result; 2 with no instrument failure: 0 honoured, 1 violated, 1 with no contract checked; 6 where Syntology's instrument failed) · 2 unverified); nothing checks that they implement this paper's method. report
DBraun/DAC-JAX mentioned on GitHubjaxSyntology: 11 samples, 9 ran (2 instrument-failed) · 2 unverified report
nvidia/bigvgan mentioned on GitHubpytorchMITSyntology: no sample linked to this paper (harvested for another paper). Syntology's graph links these samples to Multi-Scale Sub-Band Constant-Q Transform Discriminator for High-Fidelity Vocoder (arXiv:2311.14957; that paper's own run record: 11 ran (of which 0 constructed an object rather than computing a result; 8 with no instrument failure: 1 honoured, 0 violated, 7 with no contract checked; 3 where Syntology's instrument failed) · 0 unverified); nothing checks that they implement this paper's method. report
zhangxinfd/speechtokenizer mentioned on GitHubpytorchApache-2.0Syntology: 27 samples, 18 ran (8 instrument-failed) · 9 unverified report

Repository list and official/mentioned flags are the archive's, frozen 2025-07-28. Reachability, where shown, is from one Syntology probe window (2026-09-16 to 2026-09-18); repositories not probed show nothing. GitHub stars are not tracked.

A paper named beside a repository with no sample linked to this paper is shown with that paper's own run record, not this paper's: “ran” means executed on a synthesized input, not that the code is correct, and “where Syntology's instrument failed” counts failures of Syntology's instrument, not of the code. Papers are listed in arXiv-id order, at most three per repository.

Code Syntology ran Syntology

38 samples harvested: 27 ran (of which 10 constructed an object rather than computing a result; 17 with no instrument failure: 3 honoured, 3 violated, 11 with no contract checked; 10 where Syntology's instrument failed) · 11 unverified. “Where Syntology's instrument failed” counts failures of Syntology's instrument, not of the code. Syntology's record for this page has not changed since , the first build that kept a record date for it; when this build read Syntology's graph is in the build record. That date is not when the samples ran.

3ran · honoured contract
3ran · violated contract
4ran · our draft was wrong
6ran; the contract check could not run
11ran
11unverified

Licence: 0 of the 38 samples are pointer only, meaning Syntology does not serve that copy's text. This page shows no code text for any sample; each one links to its file in the repository.

Harvested from 2 repositories linked to this paper, official or community; the list below is grouped by repository, each group headed by its name and what it is to the paper. “Ran” means the sample executed on a synthesized input. It does not mean the output is correct, and nothing here reproduces the paper's results. A line beginning “fixture:” under a sample that ran gives the arguments Syntology's fixture called the sample with, each as name:type/rank/element kind (? where none was recorded), and the type of what the fixture's run returned, exactly as the harness recorded them; the arguments and the returned type were recorded by two separate runs of the same fixture. No output value is shown, and a returned type says nothing about whether the output was right. “Level recorded” gives the UTC date when Syntology's graph wrote the sample's verification level, not when the code executed; where the level was written before Syntology's graph kept that date, the row says “level recorded, date not recorded”. For 6 samples that ran, the graph records a signature this page does not show: for a class the fixture built without calling, the harness called the object with inputs it made up itself, so the recorded signature is the harness's probe, not the fixture's arguments. “Honoured” and “violated” refer to a contract Syntology drafted from the code itself. “Our draft was wrong” and “the contract check could not run” are failures of Syntology's instrument, not of the code: the sample ran on Syntology's synthesized input, and then the drafted contract did not describe it, or the check itself raised an error (RAISES), lacked a dependency (DEP_MISSING) or ran out of time (TIMEOUT); hover the label for which. A metamorphic tier, where shown, is evidence of the output's shape, not that the sample implements the method.

Each sample ends with its code_sha256. A sample's code_sha256 is Syntology's identity for that exact code. It is printed in full (16 hexadecimal characters for every sample in this build), and get_code(code_sha256="…") takes it exactly as printed. An agent calls it on Syntology's MCP service (how to connect); click an identity to copy that call.

Repository labels, per sample. official repository: The archive marks this repository official for the paper. named in the paper: The archive records that the paper mentions this repository; it is not marked official. community (archive-listed): In the archive's code links for this paper, not marked official and not recorded as mentioned in the paper. found in paper text by Syntology: Syntology found this repository in the paper's own text; whether it is the authors' implementation is not asserted. community: Not in the archive's code links for this paper; a community repository Syntology harvested. Samples are grouped by repository: a repository marked official first, then by how directly the repository is tied to the paper, then in the archive's order; never by how many ran. Licence labels name the repository's licence as recorded at harvest. “Pointer only” means Syntology does not serve that copy's text, for one of six reasons: no licence file was found; the licence was not identified; the licence is recorded as permissive but that copy's record is not marked cleared; the licence is outside the permissive list Syntology serves text under (MIT, Apache-2.0, BSD and similar); this paper's copy was not located, so its licence is not recorded; or this paper's copy is cleared, but the stored copy of the same code that Syntology serves is not. Some licences outside that list permit redistribution, such as WTFPL, and GPL-3.0 under its conditions; they are simply not on the list. Hover a licence label for the reason. File links open the file on GitHub at the default branch, which may have changed since the harvest.

A function or class name links to its page when the name occurs in the harvested code of at least 5 papers; the name is a string, not an identity. A census tag appears when Syntology's census ran implementations with that name on one shared input per bucket: it gives how many were compared, from how many papers, and how many distinct outputs they produced, and names this sample's bucket when this sample was one of them.

DBraun/DAC-JAX community (archive-listed) 11 samples, 9 ran (2 instrument-failed) · 2 unverified
QuantizedResult src/dac_jax/nn/quantize.py ran MIT (permissive) · 00d0f325d81ba7bb · report fixture: returned QuantizedResult · level recorded 2026-09-22
WNConv1d src/dac_jax/nn/quantize.py ran MIT (permissive) · 82529cf058f32742 · report fixture: returned WNConv1d · level recorded 2026-09-22
conv_to_delay src/dac_jax/nn/quantize.py ran · honoured contract fingerprinted MIT (permissive) · 72e3782101525ae7 · report fixture: signature s:int/0/int, d:int/0/int, k:int/0/int, L:int/0/int · returned int · level recorded 2026-09-22
conv_to_output_length src/dac_jax/nn/quantize.py ran · honoured contract fingerprinted MIT (permissive) · 0a1da93fdea7909d · report fixture: signature s:int/0/int, d:int/0/int, k:int/0/int, L:int/0/int · returned int · level recorded 2026-09-22
default_kernel_dilation src/dac_jax/nn/quantize.py ran · honoured contract MIT (permissive) · 2e20c64249054743 · report fixture: signature kernel_dilation:NoneType/0/? · returned int · level recorded 2026-09-22
default_kernel_size src/dac_jax/nn/quantize.py ran · violated contract fingerprinted MIT (permissive) · c9c9105f86fbbcc1 · report fixture: signature kernel_size:tuple/1/int · returned int · level recorded 2026-09-22
default_stride src/dac_jax/nn/quantize.py ran · violated contract fingerprinted MIT (permissive) · 9e520e492c31a453 · report fixture: signature strides:list/1/int · returned int · level recorded 2026-09-22
mse_loss in 73 papers
census: 8 implementations from 12 papers, 6 distinct outputs across 4 buckets, one shared input per bucket
src/dac_jax/nn/quantize.py ran; the contract check could not run (RAISES) MIT (permissive) · dff46a2b65e582bd · report fixture: signature predictions:ArrayImpl/2/float, targets:ArrayImpl/2/float, reduction:str/0/str · returned ArrayImpl · level recorded 2026-09-22
normalize in 638 papers
census: 70 implementations from 82 papers, 51 distinct outputs across 11 buckets, one shared input per bucket
src/dac_jax/nn/quantize.py ran; the contract check could not run (RAISES) fingerprinted MIT (permissive) · e29415637b6730ed · report fixture: signature x:ArrayImpl/2/float, ord:int/0/int, axis:int/0/int, eps:float/0/float · returned ArrayImpl · level recorded 2026-09-22
ResidualVectorQuantize src/dac_jax/nn/quantize.py unverified MIT (permissive) · 14c03dd20a30e62c · report
VectorQuantize in 6 papers src/dac_jax/nn/quantize.py unverified MIT (permissive) · eb961f73627c1425 · report
zhangxinfd/speechtokenizer community (archive-listed) 27 samples, 18 ran (8 instrument-failed) · 9 unverified
EuclideanCodebook in 6 papers speechtokenizer/model.py ran · metamorphic tier: deterministic Apache-2.0 (permissive) · 88ebb93ff05f77f2 · report fixture: returned EuclideanCodebook · level recorded 2026-09-22
NormConv1d speechtokenizer/model.py ran fingerprinted Apache-2.0 (permissive) · fc59f82698306aa0 · report fixture: signature x:Tensor/3/float · returned Tensor · level recorded 2026-09-24
NormConvTranspose1d speechtokenizer/model.py ran fingerprinted Apache-2.0 (permissive) · f0f69a2c5ff07a17 · report fixture: returned NormConvTranspose1d · level recorded 2026-09-22
QuantizedResult speechtokenizer/model.py ran · metamorphic tier: deterministic Apache-2.0 (permissive) · 1317f68bf0b61f08 · report fixture: returned QuantizedResult · level recorded 2026-09-22
SConv1d speechtokenizer/model.py ran · metamorphic tier: deterministic fingerprinted Apache-2.0 (permissive) · 6aee5fa73a0d5daf · report fixture: returned SConv1d · level recorded 2026-09-22
SConvTranspose1d speechtokenizer/model.py ran · metamorphic tier: deterministic fingerprinted Apache-2.0 (permissive) · 5429aefc43f5e1b4 · report fixture: returned SConvTranspose1d · level recorded 2026-09-22
SLSTM speechtokenizer/model.py ran · metamorphic tier: deterministic fingerprinted Apache-2.0 (permissive) · 46e56c21d7e192a3 · report fixture: returned SLSTM · level recorded 2026-09-22
Snake1d speechtokenizer/model.py ran · metamorphic tier: deterministic Apache-2.0 (permissive) · 27d28bc9b91d474b · report fixture: returned Snake1d · level recorded 2026-09-22
VectorQuantization speechtokenizer/model.py ran · metamorphic tier: deterministic Apache-2.0 (permissive) · 1a5bd5c497f7d3f9 · report fixture: returned VectorQuantization · level recorded 2026-09-22
apply_parametrization_norm in 7 papers speechtokenizer/model.py ran · our draft was wrong Apache-2.0 (permissive) · 7e451c435923824b · report fixture: signature module:callable/0/?, norm:str/0/str · returned Conv1d · level recorded 2026-09-22
default in 430 papers speechtokenizer/model.py ran · our draft was wrong Apache-2.0 (permissive) · 1e8f03292b45230b · report fixture: signature val:NoneType/0/?, d:str/0/str · returned str · level recorded 2026-09-22
get_extra_padding_for_conv1d in 9 papers speechtokenizer/model.py ran · our draft was wrong fingerprinted Apache-2.0 (permissive) · 5155996ee4a7f5c4 · report fixture: signature x:Tensor/3/float, kernel_size:int/0/int, stride:int/0/int, padding_total:int/0/int · returned int · level recorded 2026-09-22
get_norm_module in 6 papers speechtokenizer/model.py ran · our draft was wrong Apache-2.0 (permissive) · 6c7c136baea4aa36 · report fixture: signature module:callable/0/?, causal:bool/0/bool, norm:str/0/str · returned Identity · level recorded 2026-09-22
kmeans in 38 papers speechtokenizer/model.py ran; the contract check could not run (RAISES) · metamorphic tier: deterministic Apache-2.0 (permissive) · c26e79c6acd99d8d · report fixture: signature samples:Tensor/2/float, num_clusters:int/0/int, num_iters:int/0/int · returned tuple · level recorded 2026-09-22
laplace_smoothing in 8 papers
census: 3 implementations from 6 papers, 2 distinct outputs across 2 buckets, one shared input per bucket this sample: bucket 1 of 2, arg 1: rank 1, kind float, dtype float32
speechtokenizer/model.py ran · violated contract Apache-2.0 (permissive) · c360b5ccff704a88 · report fixture: signature x:Tensor/1/float, n_categories:int/0/int, epsilon:float/0/float · returned Tensor · level recorded 2026-09-22
sample_vectors in 15 papers
census: 3 implementations from 5 papers, 1 distinct output on one shared input this sample: bucket 1 of 1, arg 1: rank 2, kind float, dtype float32
speechtokenizer/model.py ran; the contract check could not run (RAISES) · metamorphic tier: invariant fingerprinted Apache-2.0 (permissive) · 09f10cd96592b37d · report fixture: signature samples:Tensor/2/float, num:int/0/int · returned Tensor · level recorded 2026-09-22
snake in 10 papers speechtokenizer/model.py ran; the contract check could not run (RAISES) · metamorphic tier: invariant Apache-2.0 (permissive) · f247e4748f379a76 · report fixture: returned Tensor · level recorded 2026-09-22
unpad1d speechtokenizer/model.py ran; the contract check could not run (RAISES) · metamorphic tier: invariant Apache-2.0 (permissive) · 3f2ccb02099faf04 · report fixture: signature x:Tensor/2/float, paddings:tuple/1/int · returned Tensor · level recorded 2026-09-22
ResidualVectorQuantization speechtokenizer/model.py unverified Apache-2.0 (permissive) · a4459f091cbc21cc · report
ResidualVectorQuantizer speechtokenizer/model.py unverified Apache-2.0 (permissive) · 539769016ca08bfd · report
SEANetDecoder speechtokenizer/model.py unverified Apache-2.0 (permissive) · e27b8e6c0bf19894 · report
SEANetEncoder speechtokenizer/model.py unverified Apache-2.0 (permissive) · 27c322d4a4fa97a2 · report
SEANetResnetBlock speechtokenizer/model.py unverified Apache-2.0 (permissive) · eb210115baa57d3b · report
SpeechTokenizer speechtokenizer/model.py unverified Apache-2.0 (permissive) · 707e1121f91ae8d3 · report
ema_inplace speechtokenizer/model.py unverified Apache-2.0 (permissive) · 641438f624dd1c49 · report
pad1d in 5 papers speechtokenizer/model.py unverified Apache-2.0 (permissive) · 7b7293843c2034b2 · report
uniform_init in 10 papers speechtokenizer/model.py unverified Apache-2.0 (permissive) · d30ebe7be75adcf6 · report

Method as code Syntology

What Syntology's graph records as this paper's claimed method, lifted into code: each assembly names its bricks (functions and classes), each with its own role and its own verification level. An assembly is not a check that the code implements the method correctly or reproduces the paper's results. Levels are each brick's own verification level in Syntology's graph (level 0: no recorded run), listed brick by brick and never combined into one level; “ran” means the brick executed on a synthesized input, not that its output is correct, and the instrument figure counts failures of Syntology's instrument, not of the code. Where a model read the code against the paper's claim, each assembly's label verdict is one model's judgement, kept in this page's JSON twin; it is not calibrated, so no card prints it. Siblings are assemblies lifted from this paper in the same batch for the same method: alternative readings kept side by side; none replaces or supersedes another.

2 assemblies on this paper: 2 harvested, 0 generated, listed in 1 groups (siblings lifted in one batch form one group).

2 sibling assemblies for Improved RVQGAN (harvested), lifted in the same batch on 2026-09-09: alternative readings kept side by side; none replaces or supersedes another.

harvested Assembly for Improved RVQGAN (graph method: Improved RVQGAN). 27 bricks; individually on a fixture: 18 ran (of which 8 constructed an object rather than computing a result; 10 with no instrument failure: 0 honoured, 1 violated, 9 with no contract checked; 8 where Syntology's instrument failed) · 9 unverified. The assembly has not been run end-to-end.

one of 2 siblings

  1. SpeechTokenizer entry · level 0 · unverified · 707e1121f91ae8d3
  2. ResidualVectorQuantizer model · level 0 · unverified · 539769016ca08bfd
  3. SEANetDecoder helper · level 0 · unverified · e27b8e6c0bf19894
  4. SEANetEncoder helper · level 0 · unverified · 27c322d4a4fa97a2
  5. QuantizedResult helper · level 1 · ran · 1317f68bf0b61f08
  6. ResidualVectorQuantization helper · level 0 · unverified · a4459f091cbc21cc
  7. SConv1d helper · level 1 · ran · 6aee5fa73a0d5daf
  8. SConvTranspose1d helper · level 1 · ran · 5429aefc43f5e1b4
  9. SEANetResnetBlock helper · level 0 · unverified · eb210115baa57d3b
  10. SLSTM helper · level 1 · ran · 46e56c21d7e192a3
  11. Snake1d helper · level 1 · ran · 27d28bc9b91d474b
  12. VectorQuantization helper · level 1 · ran · 1a5bd5c497f7d3f9
  13. NormConv1d helper · level 1 · ran · fc59f82698306aa0
  14. get_extra_padding_for_conv1d helper · level 1 · ran · our draft was wrong · 5155996ee4a7f5c4
  15. pad1d helper · level 0 · unverified · 7b7293843c2034b2
  16. NormConvTranspose1d helper · level 1 · ran · f0f69a2c5ff07a17
  17. unpad1d helper · level 1 · ran; the contract check could not run (RAISES) · 3f2ccb02099faf04
  18. snake helper · level 1 · ran; the contract check could not run (RAISES) · f247e4748f379a76
  19. EuclideanCodebook helper · level 1 · ran · 88ebb93ff05f77f2
  20. default helper · level 1 · ran · our draft was wrong · 1e8f03292b45230b
  21. apply_parametrization_norm helper · level 1 · ran · our draft was wrong · 7e451c435923824b
  22. get_norm_module helper · level 1 · ran · our draft was wrong · 6c7c136baea4aa36
  23. ema_inplace helper · level 0 · unverified · 641438f624dd1c49
  24. kmeans helper · level 1 · ran; the contract check could not run (RAISES) · c26e79c6acd99d8d
  25. laplace_smoothing helper · level 1 · ran · violated contract · c360b5ccff704a88
  26. sample_vectors helper · level 1 · ran; the contract check could not run (RAISES) · 09f10cd96592b37d
  27. uniform_init helper · level 0 · unverified · d30ebe7be75adcf6

harvested Assembly for Improved RVQGAN (graph method: Improved RVQGAN). 11 bricks; individually on a fixture: 9 ran (of which 2 constructed an object rather than computing a result; 7 with no instrument failure: 3 honoured, 2 violated, 2 with no contract checked; 2 where Syntology's instrument failed) · 2 unverified. The assembly has not been run end-to-end.

one of 2 siblings

  1. ResidualVectorQuantize entry · level 0 · unverified · 14c03dd20a30e62c
  2. QuantizedResult helper · level 1 · ran · 00d0f325d81ba7bb
  3. VectorQuantize helper · level 0 · unverified · eb961f73627c1425
  4. WNConv1d helper · level 1 · ran · 82529cf058f32742
  5. mse_loss loss · level 1 · ran; the contract check could not run (RAISES) · dff46a2b65e582bd
  6. normalize helper · level 1 · ran; the contract check could not run (RAISES) · e29415637b6730ed
  7. conv_to_delay helper · level 1 · ran · honoured contract · 72e3782101525ae7
  8. conv_to_output_length helper · level 1 · ran · honoured contract · 0a1da93fdea7909d
  9. default_kernel_dilation helper · level 1 · ran · honoured contract · 2e20c64249054743
  10. default_kernel_size helper · level 1 · ran · violated contract · c9c9105f86fbbcc1
  11. default_stride helper · level 1 · ran · violated contract · 9e520e492c31a453

Each brick ends with its code_sha256, Syntology's identity for that exact code; click one to copy the get_code call that returns it. For agents, on Syntology's MCP service: assembly(arxiv_id="2306.06546") returns these assemblies with each brick's role and wiring (how to connect).

Tasks

Audio CompressionAudio GenerationQuantization

Results from the paper archive 2025-07-28

No leaderboard rows for this paper in the archive.

Report a problem or propose a change · a person checks every report against the paper or source before anything changes; decisions are listed on /corrections