Methods › Sequential › Recurrent Neural Networks › WaveRNN

WaveRNN

26 papers tagged archive 2025-07-28

Introduced by Nal Kalchbrenner et al. in Efficient Neural Audio Synthesis

archive 2025-07-28 Description, source and code snippet are the archive's method entry.

WaveRNN is a single-layer recurrent neural network for audio generation that is designed efficiently predict 16-bit raw audio samples.

The overall computation in the WaveRNN is as follows (biases omitted for brevity):

𝐱ₜ = [𝐜ₜ₋₁,𝐟ₜ₋₁, 𝐜ₜ]

𝐮ₜ = σ(𝐑ᵤ𝐡ₜ₋₁ + 𝐈^*ᵤ𝐱ₜ)

𝐫ₜ = σ(𝐑ᵣ𝐡ₜ₋₁ + 𝐈^*ᵣ𝐱ₜ)

𝐞ₜ = τ(𝐫ₜ ⊙(𝐑ₑ𝐡ₜ₋₁) + 𝐈^*ₑ𝐱ₜ )

𝐡ₜ = 𝐮ₜ ·𝐡ₜ₋₁ + (1-𝐮ₜ) ·𝐞ₜ

𝐲_c, 𝐲_f = split(𝐡ₜ)

P(𝐜ₜ) = softmax(𝐎₂relu(𝐎₁𝐲_c))

P(𝐟ₜ) = softmax(𝐎₄relu(𝐎₃𝐲_f))

where the * indicates a masked matrix whereby the last coarse input 𝐜ₜ is only connected to the fine part of the states 𝐮ₜ, 𝐫ₜ, 𝐞ₜ and 𝐡ₜ and thus only affects the fine output 𝐲_f. The coarse and fine parts 𝐜ₜ and 𝐟ₜ are encoded as scalars in [0, 255] and scaled to the interval [−1, 1]. The matrix 𝐑 formed from the matrices 𝐑ᵤ, 𝐑ᵣ, 𝐑ₑ is computed as a single matrix-vector product to produce the contributions to all three gates 𝐮ₜ, mathbfrₜ and 𝐞ₜ (a variant of the GRU cell. σ and τ are the standard sigmoid and tanh non-linearities.

Each part feeds into a softmax layer over the corresponding 8 bits and the prediction of the 8 fine bits is conditioned on the 8 coarse bits. The resulting Dual Softmax layer allows for efficient prediction of 16-bit samples using two small output spaces (2 8 values each) instead of a single large output space (with 2 16 values).

PaperSource

Papers archive 2025-07-28

26 shown of 26, newest first. Repository counts are the archive's code-links table. A Syntology line states what Syntology ran from that paper's harvested code; it is per sample and not a correctness claim.

Tasks archive 2025-07-28

20 shown of 27 tasks the archive attaches to papers tagged with this method, by distinct papers. A task without a page in the catalog is plain text.

TaskPapers
Speech Synthesis14
Text to Speech11
text-to-speech11
CPU5
Text-To-Speech Synthesis4
GPU3
Decoder2
Denoising2
Face Swapping2
Transfer Learning2
Voice Conversion2
Audio Classification1
Audio Synthesis1
Automatic Speech Recognition1
Automatic Speech Recognition (ASR)1
Edge Classification1
Face Model1
Generative Adversarial Network1
Low-latency processing1
Polyphone disambiguation1

Usage over time archive 2025-07-28

Papers per year tagged with WaveRNN: 2018 to 2024, peak 7 7 0 2018: 2 papers 2018 2019: 3 papers 2019 2020: 7 papers 2020 2021: 5 papers 2021 2022: 7 papers 2022 2023: 1 paper 2023 2024: 1 paper 2024
Papers per year the archive tags with this method, by the paper's archive date (26 dated). Bars are counts, not a trend claim.

Components: the archive holds no method-to-method composition, so PwC's Components table cannot be rebuilt; the Papers list carries no Results column for the same reason (the archive does not join its leaderboard rows to method tags).

Categories archive 2025-07-28

Recurrent Neural NetworksGenerative Audio Models

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