| Paper Abstract and Keywords |
| Presentation |
2025-06-13 13:20
A Study on Improving Diffusion-Based Voice Conversion with Consistency Trajectory Models Ryuichi Hatakeyama, Toru Nakashika (UEC) SP2025-10 |
| Abstract |
(in Japanese) |
(See Japanese page) |
| (in English) |
This paper proposes a fast voice-conversion framework that replaces the diffusion decoder of DDPMVC with a Consistency Trajectory Model (CTM), a model capable of learning score estimation and trajectory-consistent jumps simultaneously.
DDPMVC achieves high conversion accuracy by coupling (i) a diffusion encoder that suppresses speaker-dependent information in the input speech and (ii) a decoder that generates the target speaker’s mel-spectrogram through many steps of reverse diffusion sampling; however, its inference time is lengthy because of those sequential steps. Although diffusion-based VC already outperforms conventional deep-learning approaches in quality, its sequential sampling still limits generation speed.
CTM can predict the end point of the diffusion trajectory directly in only a few steps while yielding higher perceptual quality than the original diffusion sampler. Motivated by this property, we investigate a voice-conversion system in which the conventional diffusion decoder is entirely replaced by CTM, enabling the generation of high-quality mel-spectrograms with dramatically fewer sampling steps. |
| Keyword |
(in Japanese) |
(See Japanese page) |
| (in English) |
Diffusion Model / DDPM / Consistency Trajectory Models / Voice Conversion / DDPMVC / VoiceGrad / / |
| Reference Info. |
IEICE Tech. Rep., vol. 125, no. 74, SP2025-10, pp. 52-56, June 2025. |
| Paper # |
SP2025-10 |
| Date of Issue |
2025-06-06 (SP) |
| ISSN |
Online edition: ISSN 2432-6380 |
Copyright and reproduction |
All rights are reserved and no part of this publication may be reproduced or transmitted in any form or by any means, electronic or mechanical, including photocopy, recording, or any information storage and retrieval system, without permission in writing from the publisher. Notwithstanding, instructors are permitted to photocopy isolated articles for noncommercial classroom use without fee. (License No.: 10GA0019/12GB0052/13GB0056/17GB0034/18GB0034) |
| Download PDF |
SP2025-10 |
| Conference Information |
| Committee |
SP IPSJ-SLP IPSJ-MUS |
| Conference Date |
2025-06-13 - 2025-06-14 |
| Place (in Japanese) |
(See Japanese page) |
| Place (in English) |
|
| Topics (in Japanese) |
(See Japanese page) |
| Topics (in English) |
|
| Paper Information |
| Registration To |
SP |
| Conference Code |
2025-06-SP-SLP-MUS |
| Language |
Japanese |
| Title (in Japanese) |
(See Japanese page) |
| Sub Title (in Japanese) |
(See Japanese page) |
| Title (in English) |
A Study on Improving Diffusion-Based Voice Conversion with Consistency Trajectory Models |
| Sub Title (in English) |
|
| Keyword(1) |
Diffusion Model |
| Keyword(2) |
DDPM |
| Keyword(3) |
Consistency Trajectory Models |
| Keyword(4) |
Voice Conversion |
| Keyword(5) |
DDPMVC |
| Keyword(6) |
VoiceGrad |
| Keyword(7) |
|
| Keyword(8) |
|
| 1st Author's Name |
Ryuichi Hatakeyama |
| 1st Author's Affiliation |
The University of Electro-Communications (UEC) |
| 2nd Author's Name |
Toru Nakashika |
| 2nd Author's Affiliation |
The University of Electro-Communications (UEC) |
| 3rd Author's Name |
|
| 3rd Author's Affiliation |
() |
| 4th Author's Name |
|
| 4th Author's Affiliation |
() |
| 5th Author's Name |
|
| 5th Author's Affiliation |
() |
| 6th Author's Name |
|
| 6th Author's Affiliation |
() |
| 7th Author's Name |
|
| 7th Author's Affiliation |
() |
| 8th Author's Name |
|
| 8th Author's Affiliation |
() |
| 9th Author's Name |
|
| 9th Author's Affiliation |
() |
| 10th Author's Name |
|
| 10th Author's Affiliation |
() |
| 11th Author's Name |
|
| 11th Author's Affiliation |
() |
| 12th Author's Name |
|
| 12th Author's Affiliation |
() |
| 13th Author's Name |
|
| 13th Author's Affiliation |
() |
| 14th Author's Name |
|
| 14th Author's Affiliation |
() |
| 15th Author's Name |
|
| 15th Author's Affiliation |
() |
| 16th Author's Name |
|
| 16th Author's Affiliation |
() |
| 17th Author's Name |
|
| 17th Author's Affiliation |
() |
| 18th Author's Name |
|
| 18th Author's Affiliation |
() |
| 19th Author's Name |
|
| 19th Author's Affiliation |
() |
| 20th Author's Name |
|
| 20th Author's Affiliation |
() |
| 21st Author's Name |
|
| 21st Author's Affiliation |
() |
| 22nd Author's Name |
|
| 22nd Author's Affiliation |
() |
| 23rd Author's Name |
|
| 23rd Author's Affiliation |
() |
| 24th Author's Name |
|
| 24th Author's Affiliation |
() |
| 25th Author's Name |
|
| 25th Author's Affiliation |
() |
| 26th Author's Name |
/ / |
| 26th Author's Affiliation |
()
() |
| 27th Author's Name |
/ / |
| 27th Author's Affiliation |
()
() |
| 28th Author's Name |
/ / |
| 28th Author's Affiliation |
()
() |
| 29th Author's Name |
/ / |
| 29th Author's Affiliation |
()
() |
| 30th Author's Name |
/ / |
| 30th Author's Affiliation |
()
() |
| 31st Author's Name |
/ / |
| 31st Author's Affiliation |
()
() |
| 32nd Author's Name |
/ / |
| 32nd Author's Affiliation |
()
() |
| 33rd Author's Name |
/ / |
| 33rd Author's Affiliation |
()
() |
| 34th Author's Name |
/ / |
| 34th Author's Affiliation |
()
() |
| 35th Author's Name |
/ / |
| 35th Author's Affiliation |
()
() |
| 36th Author's Name |
/ / |
| 36th Author's Affiliation |
()
() |
| Speaker |
Author-1 |
| Date Time |
2025-06-13 13:20:00 |
| Presentation Time |
140 minutes |
| Registration for |
SP |
| Paper # |
SP2025-10 |
| Volume (vol) |
vol.125 |
| Number (no) |
no.74 |
| Page |
pp.52-56 |
| #Pages |
5 |
| Date of Issue |
2025-06-06 (SP) |