{"url":"/method/dv3-convolution-block","slug":"dv3-convolution-block","name":"DV3 Convolution Block","full_name":"DV3 Convolution Block","full_name_withheld":false,"description_markdown":"**DV3 Convolution Block** is a convolutional block used for the [Deep Voice 3](https://paperswithcode.com/method/deep-voice-3) text-to-speech architecture. It consists of a 1-D [convolution](https://paperswithcode.com/method/convolution) with a gated linear unit and a [residual connection](https://paperswithcode.com/method/residual-connection). In the Figure, $c$ denotes the dimensionality of the input. The convolution output of size $2 \\cdot c$ is split into equal-sized portions: the gate vector and the input vector. A scaling factor $\\sqrt{0.5}$ is used to ensure that we preserve the input variance early in training. The gated linear unit provides a linear path for the gradient flow, which alleviates the vanishing gradient issue for stacked convolution blocks while retaining non-linearity. To introduce speaker-dependent control, a speaker-dependent embedding is added as a bias to the convolution filter output, after a softsign function. The authors use the softsign nonlinearity because it limits the range of the output while also avoiding the saturation problem that exponential based nonlinearities sometimes exhibit. Convolution filter weights are initialized with zero-mean and unit-variance activations throughout the entire network.","description_state":"present","introduced_year":null,"introduced_by":{"title":null,"paper":null,"first_author":null,"n_authors":0,"url_abs":null,"archive_paper_url":null},"source":{"url":"http://arxiv.org/abs/1710.07654v3","title":"Deep Voice 3: Scaling Text-to-Speech with Convolutional Sequence Learning","url_on_a_paper_host":true},"code_snippet_url":"https://github.com/r9y9/deepvoice3_pytorch/blob/897f31e57eb6ec2f0cafa8dc62968e60f6a96407/deepvoice3_pytorch/modules.py#L112","code_snippet_url_on_a_code_host":true,"categories":[{"area":"Audio","area_id":"audio","collection":"Audio Model Blocks","url":"/methods/category/audio-model-blocks","pwc_aliases":[]}],"n_papers_tagged":9,"archive_num_papers":null,"papers_newest_first":[{"paper":"/paper/clarinet-a-music-retrieval-system","title":"Clarinet: A Music Retrieval System","date":"2022-10-23","arxiv_id":"2210.12648","n_code_links":1,"syntology":null},{"paper":"/paper/learning-from-a-complementary-label-source","title":"Learning from a Complementary-label Source Domain: Theory and Algorithms","date":"2020-08-04","arxiv_id":"2008.01454","n_code_links":1,"syntology":{"ran":3,"of":4,"unverified":1,"pointer_only":4}},{"paper":"/paper/clarinet-a-one-step-approach-towards-budget","title":"Clarinet: A One-step Approach Towards Budget-friendly Unsupervised Domain Adaptation","date":"2020-07-29","arxiv_id":"2007.14612","n_code_links":1,"syntology":{"ran":3,"of":4,"unverified":1,"pointer_only":4}},{"paper":null,"title":"Parallel Neural Text-to-Speech","date":"2020-01-01","arxiv_id":null,"n_code_links":0,"syntology":null},{"paper":null,"title":"Multi-Speaker End-to-End Speech Synthesis","date":"2019-07-09","arxiv_id":"1907.04462","n_code_links":0,"syntology":null},{"paper":"/paper/parallel-neural-text-to-speech","title":"Non-Autoregressive Neural Text-to-Speech","date":"2019-05-21","arxiv_id":"1905.08459","n_code_links":2,"syntology":{"ran":2,"of":7,"unverified":5,"pointer_only":0}},{"paper":null,"title":"Neural source-filter waveform models for statistical parametric speech synthesis","date":"2019-04-27","arxiv_id":"1904.12088","n_code_links":0,"syntology":null},{"paper":"/paper/clarinet-parallel-wave-generation-in-end-to","title":"ClariNet: Parallel Wave Generation in End-to-End Text-to-Speech","date":"2018-07-19","arxiv_id":"1807.07281","n_code_links":5,"syntology":{"ran":1,"of":13,"unverified":12,"pointer_only":0}},{"paper":"/paper/deep-voice-3-scaling-text-to-speech-with","title":"Deep Voice 3: Scaling Text-to-Speech with Convolutional Sequence Learning","date":"2017-10-20","arxiv_id":"1710.07654","n_code_links":7,"syntology":{"ran":1,"of":1,"unverified":0,"pointer_only":1}}],"papers_shown":9,"tasks":[{"task":"/task/speech-synthesis","name":"Speech Synthesis","papers":4},{"task":"/task/text-to-speech","name":"Text to Speech","papers":4},{"task":"/task/text-to-speech-1","name":"text-to-speech","papers":4},{"task":"/task/domain-adaptation","name":"Domain Adaptation","papers":2},{"task":"/task/unsupervised-domain-adaptation","name":"Unsupervised Domain Adaptation","papers":2},{"task":null,"name":"GPU","papers":1},{"task":"/task/melody-extraction","name":"Melody Extraction","papers":1},{"task":"/task/retrieval","name":"Retrieval","papers":1},{"task":"/task/text-to-speech-synthesis","name":"Text-To-Speech Synthesis","papers":1}],"tasks_shown":9,"n_tasks":9,"usage_by_year":[{"year":"2017","papers":1},{"year":"2018","papers":1},{"year":"2019","papers":3},{"year":"2020","papers":3},{"year":"2022","papers":1}],"row_source":"embedded","archive":{"source":"pwc-archive (Hugging Face), CC BY-SA 4.0","snapshot":"2025-07-28","archive_url":"https://paperswithcode.com/method/dv3-convolution-block"},"syntology_read_at":"2026-09-24T18:15:14+00:00"}