read - 随笔分类 - NoNoe

歌声转换SVC主流方法原理剖析4 — ReFlow-VAE-SVC

摘要：pre 本文SVC指的是歌声转换（Singing Voice Conversion (SVC)），例如常见且开源的 So-VITS-SVC, RVC, DDSP-SVC 关键词：歌声转换、声音克隆、AI翻唱本来是不打算写ReFlow-VAE-SVC的，不过实在是对名字里面那个VAE很在意，而且由于阅读全文

posted @ 2025-12-03 22:31 NoNoe 阅读(65) 评论(0) 推荐(0)

歌声转换SVC主流方法原理剖析3 — So-VITS-SVC

摘要：pre 本文SVC指的是歌声转换（Singing Voice Conversion (SVC)），例如常见且开源的 So-VITS-SVC, RVC, DDSP-SVC 关键词：歌声转换、声音克隆、AI翻唱 DDSP-SVC训练快，但总是有音色泄漏。RIFT-SVC训练会慢上许多，效果略好。 So- 阅读全文

posted @ 2025-12-02 23:15 NoNoe 阅读(75) 评论(0) 推荐(0)

歌声转换SVC主流方法原理剖析2 — RIFT-SVC

摘要：pre 本文SVC指的是歌声转换（Singing Voice Conversion (SVC)），例如常见且开源的 So-VITS-SVC, RVC, DDSP-SVC 关键词：歌声转换、声音克隆、AI翻唱 DDSP-SVC训练快，但总是有音色泄漏，瞎改了几下似乎也没啥帮助。于是试试RIFT-SVC 阅读全文

posted @ 2025-11-09 20:22 NoNoe 阅读(100) 评论(0) 推荐(0)

歌声转换SVC主流方法原理剖析1 — DDSP-SVC

摘要：pre 本文SVC指的是歌声转换（Singing Voice Conversion (SVC)），例如常见且开源的 So-VITS-SVC, RVC, DDSP-SVC 关键词：歌声转换、声音克隆、AI翻唱最早在23年刷到了惠惠的冬之花翻唱，惊为天人，一直对这块很感兴趣，奈何当时有其他研究，平时时阅读全文

posted @ 2025-10-29 22:14 NoNoe 阅读(126) 评论(0) 推荐(0)

[论文阅读] IF-Font@ Ideographic Description Sequence-Following Font Generation

摘要：Pre title: IF-Font: Ideographic Description Sequence-Following Font Generation source: NeurIPS 2024 paper: https://proceedings.neurips.cc/paper_files/ 阅读全文

posted @ 2025-03-20 13:24 NoNoe 阅读(786) 评论(4) 推荐(1)

[论文速览] 一些向量量化的相关工作

摘要：Pre 想认真整理却没时间，很无奈，大概就这样吧 Zero-Shot Text-to-Image Generation (DALL-E) code https://github.com/openai/DALL-E Idea 提出 dVAE 将离散采样问题放松为连续近似，VQ-VAE迫使模型在所有情况阅读全文

posted @ 2024-12-30 19:21 NoNoe 阅读(329) 评论(0) 推荐(0)

[论文阅读] Few shot font generation via transferring similarity guided global style and quantization local style

摘要：Pre title: Few shot font generation via transferring similarity guided global style and quantization local style accepted: ICCV 2023 paper: https://ar 阅读全文

posted @ 2024-12-30 19:20 NoNoe 阅读(325) 评论(2) 推荐(0)

[论文速览] Language Model Beats Diffusion - Tokenizer is Key to Visual Generation

摘要：Pre title: Language Model Beats Diffusion - Tokenizer is Key to Visual Generation accepted: ICLR 2024 paper: https://arxiv.org/abs/2310.05737 code: no 阅读全文

posted @ 2024-12-30 19:17 NoNoe 阅读(270) 评论(0) 推荐(1)

[论文速览] Vector Quantized Image-to-Image Translation

摘要：Pre title: Vector Quantized Image-to-Image Translation accepted: ECCV 2022 paper: https://arxiv.org/abs/2207.13286 code: https://github.com/cyj407/VQ- 阅读全文

posted @ 2024-12-30 19:10 NoNoe 阅读(219) 评论(2) 推荐(0)

[论文阅读] VQ-Font@ Few-Shot Font Generation with Structure-Aware Enhancement and Quantization

摘要：Pre title: VQ-Font: Few-Shot Font Generation with Structure-Aware Enhancement and Quantization accepted: arXiv 2023 paper: https://arxiv.org/abs/2308. 阅读全文

posted @ 2024-12-12 22:26 NoNoe 阅读(379) 评论(0) 推荐(0)

[论文阅读] Radical Analysis Network for Zero-Shot Learning in Printed Chinese Character Recognition

摘要：Pre title: Radical Analysis Network for Zero-Shot Learning in Printed Chinese Character Recognition accepted: ICME 2018 paper: https://arxiv.org/abs/1 阅读全文

posted @ 2024-12-12 22:23 NoNoe 阅读(183) 评论(0) 推荐(0)

[论文阅读] Vector-quantized Image Modeling with Improved VQGAN

摘要：Pre title: Vector-quantized Image Modeling with Improved VQGAN accepted: ICLR 2022 paper: https://arxiv.org/abs/2110.04627 code: https://github.com/th 阅读全文

posted @ 2024-12-05 13:02 NoNoe 阅读(398) 评论(0) 推荐(0)

[论文阅读] Breaking the Representation Bottleneck of Chinese Characters{colon}Neural Machine Translation with Stroke Sequence Modeling

摘要：Pre title: Breaking the Representation Bottleneck of Chinese Characters:Neural Machine Translation with Stroke Sequence Modeling accepted: EMNLP 2022 阅读全文

posted @ 2024-12-04 14:49 NoNoe 阅读(147) 评论(0) 推荐(0)

[论文速览] AE,VAE,VQ-VAE,VQ-GAN,FSQ

摘要：Pre ref: 《An Introduction to Autoencoders》 ref: https://zhuanlan.zhihu.com/p/388620573 ref: https://www.spaces.ac.cn/archives/5253 ref: https://zhuanl 阅读全文

posted @ 2024-12-04 14:48 NoNoe 阅读(1204) 评论(0) 推荐(1)

[论文阅读] Drawing and Recognizing Chinese Characters with Recurrent Neural Network

摘要：Pre title: Drawing and Recognizing Chinese Characters with Recurrent Neural Network source: TPAMI 2018 paper: https://arxiv.org/abs/1606.06539 code: h 阅读全文

posted @ 2024-07-07 17:07 NoNoe 阅读(453) 评论(0) 推荐(0)

[论文阅读] Calligraphy Font Generation via Explicitly Modeling Location-Aware Glyph Component Deformations

摘要：Pre title: Calligraphy Font Generation via Explicitly Modeling Location-Aware Glyph Component Deformations source: TMM 2023 paper: https://ieeexplore. 阅读全文

posted @ 2024-07-01 22:38 NoNoe 阅读(300) 评论(0) 推荐(1)

[论文阅读] BBDM@ Image-to-Image Translation With Brownian Bridge Diffusion Models

摘要：Pre title: BBDM: Image-to-Image Translation With Brownian Bridge Diffusion Models source: CVPR 2023 paper: https://arxiv.org/abs/2205.07680 code: http 阅读全文

posted @ 2024-06-18 16:47 NoNoe 阅读(909) 评论(0) 推荐(0)

[论文速览] Small-scale proxies for large-scale Transformer training instabilities

摘要：Pre title: Small-scale proxies for large-scale Transformer training instabilities source: ICLR 2024 paper: https://arxiv.org/abs/2309.14322 code: ref: 阅读全文

posted @ 2024-06-18 16:47 NoNoe 阅读(168) 评论(0) 推荐(0)

[论文速览] DualVector@ Unsupervised Vector Font Synthesis with Dual-Part Representation

摘要：Pre title: DualVector: Unsupervised Vector Font Synthesis with Dual-Part Representation accepted: CVPR2023 paper: https://arxiv.org/abs/2305.10462 cod 阅读全文

posted @ 2024-06-02 17:34 NoNoe 阅读(215) 评论(0) 推荐(0)

[论文阅读] Design and Development of a Framework For Stroke-Based Handwritten Gujarati Font Generation

摘要：1. Pre title: Design and Development of a Framework For Stroke-Based Handwritten Gujarati Font Generation source: arXiv2024 paper: https://arxiv.org/a 阅读全文

posted @ 2024-06-01 17:51 NoNoe 阅读(176) 评论(0) 推荐(0)

心有所向，日复一日，必有精进

羽ばたき方を忘れたって、飛んでる夢を見る

随笔分类 - read

公告