Route-DMD: Distribution Matching Distillation with Task-Routed Mechanism for Speech Foundation Models
cs.SD
Submitted: 2026-11-05
Updated: 2026-11-05
Code: https://github.com/iHateTheWorld555/AuK-RouteDMD
Project page: https://route-dmd.github.io
Terminology
Sources
- Z-Image: An Efficient Image Generation Foundation Model with Single-Stream Diffusion Transformer
- MOVA: Towards Scalable and Synchronized Video-Audio Generation
- Wan: Open and Advanced Large-Scale Video Generative Models
- MegaTTS 3: Sparse Alignment Enhanced Latent Diffusion Transformer for Zero-Shot Speech Synthesis
- AuK Technical Report: An Open-Source Foundational Model for Speech Generation and Editing
- Mean Flows for One-step Generative Modeling
- Decoupled DMD: CFG Augmentation as the Spear, Distribution Matching as the Shield
- SplitMeanFlow: Interval Splitting Consistency in Few-Step Generative Modeling
- DMOSpeech 2: Reinforcement Learning for Duration Prediction in Metric-Optimized Speech Synthesis
- Rethinking Training Targets, Architectures and Data Quality for Universal Speech Enhancement
- Ming-UniAudio: Speech LLM for Joint Understanding, Generation and Editing with Unified Representation
- Step-Audio-EditX Technical Report
- QuarkAudio Technical Report
- Towards Flow-Matching-based TTS without Classifier-Free Guidance
- Flow Matching for Generative Modeling
- Qwen3-TTS Technical Report
- VoxCPM2 Technical Report
- DistillAlign: Coordinating Mode Covering and Mode Seeking in Autoregressive Video Distillation
- Seed-TTS: A Family of High-Quality Versatile Speech Generation Models
- FunASR: A Fundamental End-to-End Speech Recognition Toolkit
Related papers
- Speech-DRAME: A Framework for Human-Aligned Benchmarks in Speech Role-Play
- MuLaCover: Flexible Cover Song Generation via Symbolic Tonal Control
- Optimising Neural Speech Codecs for 300bps Communication using Reinforcement Learning
- Few-Shot Open-Set Audio Classification via Transductive Prototype Refinement and Class Logit Enhancement
- Spectral Masking and Interpolation Attack (SMIA): A Black-box Adversarial Attack against Voice Authentication and Anti-Spoofing Systems
- Remix the Timbre: Diffusion-Based Style Transfer Across Polyphonic Stems