SYNTOLOGY HomeExplorerAtlasCodeMethodologyAboutDevelopersFeedPricing
Paper · 2203.09690 · ICML · 2022

A 3 T: Alignment-Aware Acoustic and Text Pretraining for Speech Synthesis and Editing

Junkun Chen, He Bai, Renjie Zheng, Xintong Li, Mingbo Ma, Liang Huang

arXiv · PDF · Open in the Atlas

Code that ran

We lifted 7 functions out of this paper's own repositories and ran 2 of them in a sandbox. "Ran" means the function executed on a synthesized input and returned a value. It is not a reproduction of the paper's results.

RepositoryRoleRan
PaddlePaddle/PaddleSpeech canonical 1 of 3
richardbaihe/a3t pwc_unofficial 1 of 4
FunctionStatusWhere it lives
parse_infofile Ran richardbaihe/a3t/aggregate_output/sedit_decode.py
code served (permissive licence) · get_code("eaea9b1425847241")
simple_repr Ran PaddlePaddle/PaddleSpeech/paddlespeech/audiotools/core/_julius.py
code served (permissive licence) · get_code("7ac28de669f54554")
adadelta_eps_decay Not yet run richardbaihe/a3t/espnet/asr/asr_utils.py
code served (permissive licence) · get_code("bb3665ddb6b8dd2d")
adam_lr_decay Not yet run richardbaihe/a3t/espnet/asr/asr_utils.py
code served (permissive licence) · get_code("1fa5ca494ed9750a")
format_figure Not yet run PaddlePaddle/PaddleSpeech/paddlespeech/audiotools/core/display.py
code served (permissive licence) · get_code("965fb4d02c97778c")
r128stats Not yet run PaddlePaddle/PaddleSpeech/paddlespeech/audiotools/core/ffmpeg.py
code served (permissive licence) · get_code("4d1f42e7242859a7")
restore_snapshot Not yet run richardbaihe/a3t/espnet/asr/asr_utils.py
code served (permissive licence) · get_code("96081a60369c59e0")

Repositories linked to this paper

Some links come from the archived Papers with Code dataset (CC BY-SA 4.0): attribution and licence.

Abstract

Recently, speech representation learning has improved many speech-related tasks such as speech recognition, speech classification, and speech-totext translation. However, all the above tasks are in the direction of speech understanding, but for the inverse direction, speech synthesis, the potential of representation learning is yet to be realized, due to the challenging nature of generating high-quality speech. To address this problem, we propose our framework, Alignment-Aware Acoustic-Text Pretraining (A 3 T), which reconstructs masked acoustic signals with text input and acoustic-text alignment during training. In this way, the pretrained model can generate high quality reconstructed spectrogram, which can be applied to the speech editing and unseen speaker TTS directly. Experiments show A 3 T outperforms SOTA models on speech editing, and improves multi-speaker speech synthesis without the external speaker verification model. 1

For agents

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

get_harvested_code_for_paper("2203.09690")
get_code_for_paper("2203.09690")
have("2203.09690")

Connect an agent — have() is free.