Hierarchical Codec Diffusion for Video-to-Speech Generation

Hugging Face Daily Papers Papers

Summary

HiCoDiT is a novel Hierarchical Codec Diffusion Transformer for video-to-speech generation that leverages the hierarchical structure of RVQ-based codec discrete speech tokens, using coarse-to-fine conditioning with dual-scale normalization to achieve strong audio-visual alignment.

Video-to-Speech (VTS) generation aims to synthesize speech from a silent video without auditory signals. However, existing VTS methods disregard the hierarchical nature of speech, which spans coarse speaker-aware semantics to fine-grained prosodic details. This oversight hinders direct alignment between visual and speech features at specific hierarchical levels during property matching. In this paper, leveraging the hierarchical structure of Residual Vector Quantization (RVQ)-based codec, we propose HiCoDiT, a novel Hierarchical Codec Diffusion Transformer that exploits the inherent hierarchy of discrete speech tokens to achieve strong audio-visual alignment. Specifically, since lower-level tokens encode coarse speaker-aware semantics and higher-level tokens capture fine-grained prosody, HiCoDiT employs low-level and high-level blocks to generate tokens at different levels. The low-level blocks condition on lip-synchronized motion and facial identity to capture speaker-aware content, while the high-level blocks use facial expression to modulate prosodic dynamics. Finally, to enable more effective coarse-to-fine conditioning, we propose a dual-scale adaptive instance layer normalization that jointly captures global vocal style through channel-wise normalization and local prosody dynamics through temporal-wise normalization. Extensive experiments demonstrate that HiCoDiT outperforms baselines in fidelity and expressiveness, highlighting the potential of discrete modelling for VTS. The code and speech demo are both available at https://github.com/Jiaxin-Ye/HiCoDiT.
Original Article
View Cached Full Text

Cached at: 04/21/26, 07:21 AM

Paper page - Hierarchical Codec Diffusion for Video-to-Speech Generation

Source: https://huggingface.co/papers/2604.15923

Abstract

HiCoDiT generates speech from videos by leveraging the hierarchical structure of discrete speech tokens, achieving better audio-visual alignment through coarse-to-fine conditioning with dual-scale normalization.

Video-to-Speech (VTS) generation aims to synthesize speech from a silent video without auditory signals. However, existing VTS methods disregard the hierarchical nature of speech, which spans coarsespeaker-aware semanticsto fine-grained prosodic details. This oversight hinders direct alignment between visual and speech features at specific hierarchical levels during property matching. In this paper, leveraging the hierarchical structure ofResidual Vector Quantization(RVQ)-basedcodec, we propose HiCoDiT, a novel HierarchicalCodecDiffusion Transformerthat exploits the inherent hierarchy ofdiscrete speech tokensto achieve strongaudio-visual alignment. Specifically, since lower-level tokens encode coarsespeaker-aware semanticsand higher-level tokens capture fine-grained prosody, HiCoDiT employs low-level and high-level blocks to generate tokens at different levels. The low-level blocks condition onlip-synchronized motionandfacial identityto capture speaker-aware content, while the high-level blocks usefacial expressionto modulateprosodic dynamics. Finally, to enable more effectivecoarse-to-fine conditioning, we propose adual-scale adaptive instance layer normalizationthat jointly captures global vocal style through channel-wise normalization and local prosody dynamics through temporal-wise normalization. Extensive experiments demonstrate that HiCoDiT outperforms baselines in fidelity and expressiveness, highlighting the potential of discrete modelling for VTS. The code and speech demo are both available at https://github.com/Jiaxin-Ye/HiCoDiT.

View arXiv pageView PDFAdd to collection

Get this paper in your agent:

hf papers read 2604\.15923

Don’t have the latest CLI?curl \-LsSf https://hf\.co/cli/install\.sh \| bash

Models citing this paper0

No model linking this paper

Cite arxiv.org/abs/2604.15923 in a model README.md to link it from this page.

Datasets citing this paper0

No dataset linking this paper

Cite arxiv.org/abs/2604.15923 in a dataset README.md to link it from this page.

Spaces citing this paper0

No Space linking this paper

Cite arxiv.org/abs/2604.15923 in a Space README.md to link it from this page.

Collections including this paper0

No Collection including this paper

Add this paper to acollectionto link it from this page.

Similar Articles

CRoCoDiL: Continuous and Robust Conditioned Diffusion for Language

arXiv cs.CL

CRoCoDiL proposes a continuous and robust conditioned diffusion approach for language that shifts masked diffusion models into a continuous semantic space, achieving superior generation quality and 10x faster sampling speeds compared to discrete methods like LLaDA.