AI Paper Atlas

Research map / Image generation / Autoregressive and latent generation

A Vector Quantized Approach for Text to Speech Synthesis on Real-World Spontaneous Speech

Li-Wei Chen, Shinji Watanabe, Alexander Rudnicky

AAAI 2023 · Image generation › Autoregressive and latent generation · approach: Neural text-to-speech synthesis · 27 citations (OpenAlex)

Read the paperCheck your idea against 137,217 papers

AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.

Measured effects reported in the paper

What was changedWhereMeasuredEffect
Replace one codebook with four codebooks (ablation)MQTTS speech quantizer · VoxCeleb test setRCER (%): 56.5 -> 19.7Improved
Increase one codebook from 1024 to 65536 codes (ablation)MQTTS speech quantizer · VoxCeleb test setRCER (%): 56.5 -> 59.9No clear effect
Increase codebook groups from four to eight (ablation)MQTTS speech quantizer · VoxCeleb test setRCER (%): 19.7 -> 14.2Improved
Scale model size from approximately 40M to approximately 200M parameters (ablation)MQTTS · GigaSpeech real-world speechWER (%): 24.9 -> 22.2Improved
Scale model size from approximately 40M to approximately 200M parameters (ablation)MQTTS · GigaSpeech real-world speechMOS-N: 3.79$\pm$0.06 -> 3.89$\pm$0.06Improved
Remove monotonic alignment (ablation)MQTTS, approximately 100M · GigaSpeech real-world speechWER (%): 22.3 -> 34.3Hurt
Add monotonic alignment (ablation)MQTTS, approximately 100M · GigaSpeech real-world speechMOS-N: 3.76$\pm$0.06 -> 3.87$\pm$0.06Improved
Replace the sub-decoder with the same number of linear layers (ablation)MQTTS, approximately 100M · GigaSpeech real-world speechP-FID: 8.58 -> 12.12Hurt
Add the sub-decoder for conditional code sampling (ablation)MQTTS, approximately 100M · GigaSpeech real-world speechMOS-N: 3.75$\pm$0.07 -> 3.87$\pm$0.06Improved
Replace attention alignment with a duration predictor (ablation)MQTTS, approximately 100M · GigaSpeech real-world speechWER (%): 22.3 -> 53.2Hurt
Use an autoregressive multi-code model instead of VITSMQTTS and VITS, approximately 100M · GigaSpeech real-world speechP-FID: 12.04 -> 8.58Improved
Use MQTTS instead of VITSMQTTS and VITS, approximately 100M · GigaSpeech real-world speechMCD: 8.73$\pm$1.19 -> 10.22$\pm$1.68Hurt

Closest papers by meaning

Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.

Open scime draft

Get the next scime Dispatch and a monthly field update

Free. We send a confirmation link first; every email has a one-click unsubscribe. Privacy