按年被提到
20162017 · 2 次2026
视觉、语音与图像生成›语音与音频生成·arXiv:1609.03499 · cs.SD

WaveNet: A Generative Model for Raw Audio

WaveNet:原始音频的生成模型
Aaron van den Oord, Sander Dieleman, Heiga Zen, Karen Simonyan, Oriol Vinyals, Alex Graves, Nal Kalchbrenner, Andrew Senior, Koray Kavukcuoglu · 2016

WaveNet:逐采样点自回归生成原始音频波形的深度网络

开始阅读 ↗PDFalphaXiv 讨论
被推荐 · 6 人首次 2017-03-08 · 最近 2026-01-16
6不同的人
6活跃月份
6讨论串
4,901被引(OpenAlex)

WaveNet 是一个直接生成原始音频波形的深度神经网络,完全概率化且自回归,每个采样点的预测分布取决于此前所有采样点,并且能在每秒数万采样点的数据上高效训练。用于语音合成时,英语和普通话的自然度均被人类听众评为显著优于当时最好的参数式和拼接式系统;单个模型可通过说话人条件切换不同说话人,训练于音乐时也能生成逼真的片段,还可用作判别模型做音素识别。

HN 为什么提它

评论者把它当作神经语音合成的里程碑,称赞其保真度和算法的简洁,同时也指出它一次只能看到约 300 毫秒的上下文,难以学到长程结构。

方法难度 4 / 5约 45 分钟经得起时间
读之前最好知道
1自回归模型(autoregressive model)
2卷积神经网络与空洞卷积(dilated convolution)
3文本转语音(TTS)基础
4音频采样与波形表示
5概率生成模型

HN 书评

推荐在前,同类按时间 · 每篇最多保留 20 条,统计用全部提及
推荐kastnerkyle · 2017-03

“It is still surprising to me how good the audio models from WaveNet and DeepVoice are (such as the sample you show)”

称 WaveNet 和 DeepVoice 的音频模型之好仍令人惊讶
Baidu Deep Voice Explained: Part 1 – the Inference Pipeline原评论 ↗
推荐jdietrich · 2018-05

“It represents a huge advance in text-to-speech fidelity, using a remarkably straightforward algorithm.”

称它用相当简单的算法大幅提升了 text-to-speech 保真度
Today’s dominant approach to A.I. has not worked out原评论 ↗
推荐zip6 · 2019-04

“you should take a look at WaveNet if you haven't already”

建议没看过的人去看看 WaveNet
Ask HN: What is the state of the art in text to speech?原评论 ↗
批评ml_thoughts · 2017-10

“The compromise is Wavenet, which can only act on a context of about ~300 ms at a time [2], which more or less precludes learning long-range structure”

Finding Waldo Using Semantic Segmentation and Tiramisu原评论 ↗
引用rryan · 2022-08

“Try an autoregressive model of the joint probability distribution of the sine wave timepoints -- like WaveNet [1].”

建议像 WaveNet 那样用自回归模型建模正弦波时间点的联合分布
Show HN: Nevermind XOR – Deep Learning Has an Issue with Sin原评论 ↗
引用rhdunn · 2026-01

“Using neural nets (machine learning) to train TTS voices has been around a long time.”

指出用神经网络训练 TTS 语音由来已久
Pocket TTS: A high quality TTS that gives your CPU a voice原评论 ↗