Search arXiv⌕ Search
Home/Archive/arXiv

arXiv · 2610.27109

Beyond Two Tracks: Speaker-Anchored Full-Duplex Dialogue Recovery

Yuxuan Chen·Yang Xiao·Jack Sun·Chak Sun·Brick Li·Sam
View original source ↗Read source document ↗Export metadata ↓

Abstract

Spoken interaction is moving from alternating turns to full-duplex conversation, where an agent listens while speaking. Learning this coordination requires natural dialogue with persistent speaker-resolved tracks, a data requirement that remains difficult to satisfy at scale. Generative recovery offers a route from ordinary recordings, but recovering two voices is not enough: their track assignments can drift as the conversation unfolds. We address this missing link with speaker-anchored routing before diffusion: a 1.84M-parameter module matches the separator's coarse streams to fixed enrollment identities and orders them before a frozen generator reconstructs the speech, without retraining the generator or altering the reconstructed audio. On 25-minute remote conversations, routing cuts wrong-track time from 43.9% to 4.1%, below the 6.2% of matched-enrollment waveform re-identification, with its largest advantage when one participant holds the floor (17.8% versus 28.6%), and raises fixed-track activity F1 from 0.46 to 0.84 while per-chunk reconstruction quality is unchanged. The same router transfers to unseen face-to-face and telephone recordings, and a margin-gated verification step brings their wrong-track time to 0.9-1.7% while embedding 20-36% of the output, where full re-identification embeds all of it. Recovered voices thus become the participant-consistent tracks that full-duplex dialogue modeling needs. Audio examples: https://elpsykongloo.github.io/beyond-two-tracks/

Explore related subjects

cs.SDcs.CLeess.AS

Keep this discovery

Explore connections, maps & timelines

BibTeXRIS

Yuxuan Chen, Yang Xiao, Jack Sun, Chak Sun, Brick Li, Sam. 2026-10-19. Beyond Two Tracks: Speaker-Anchored Full-Duplex Dialogue Recovery. https://arxiv.org/abs/2610.27109

Cite the original work for its findings. Save a collection to share your selection of sources.

KEEP EXPLORING

Related papers

arXiv2026

FFMT: Frequency-Factorized Multi-Rate Tokenization for Music-to-Dance Generation

Dance motion combines slowly varying body configurations with rapid local movements. We investigate how explicit temporal-frequency decomposition can guide the temporal resolution and codebook capacity of discrete motion representations. We introduce frequency-factorized motion tokenization (FFMT), which separates root-relative motion into a low-frequency choreographic scaffold and a complementary high-frequency residual. FFMT uses a lower token rate and a smaller codebook for the low-frequency component, and a higher token rate and a larger codebook for the high-frequency residual. This representation further induces a plan-then-articulate autoregressive factorization: low-frequency tokens establish coarse choreography, while high-frequency tokens realize local motion details conditioned on the generated plan and music. Experiments on AIST++ and FineDance show strong distributional fidelity while preserving realistic diversity and music--motion synchronization. Ablations consistently support frequency factorization, asymmetric temporal allocation, and low-to-high conditioning.

cs.SD↗
arXiv2026

Adaptive and High-Fidelity Matching Algorithm for Retrieval-Based Accent Conversion

Accent conversion (AC) aims to transform accent-related acoustic characteristics of the source speech to a target accent while preserving linguistic content and speaker identity. Unlike conventional generative AC models, the retrieval-based ones, which employ k-nearest neighbors (kNN) to directly retrieve acoustic features from target-accent speech and so do not need source-accent speech during training, are more effective. However, existing retrieval-based AC frameworks face three issues: limited intelligibility due to incomplete phoneme coverage from sequential matching pool construction, limited adaptability to source speech variations, and low spectral continuity from coarse-grained style modulation. To address these limitations, we propose Anisotropy-Aware Adaptive k-Nearest Neighbor Accent Conversion (AA-kNN-AC), an enhanced retrieval-based AC model. First, we design a phoneme coverage-based matching pool construction strategy, employing greedy selection with diversity-aware scoring to achieve balanced and near-complete phonetic representation under limited target accent data. Second, we propose an anisotropy-aware adaptive kNN retrieval mechanism that dynamically determines the neighborhood size according to the singular value decomposition (SVD)-based anisotropy analysis, enhancing adaptability to source speech variations. Third, we present a progressive Adaptive Instance Normalization (AdaIN)-based fine-grained decoder by integrating temporal up-sampling within AdaIN blocks, enhancing spectral continuity of the converted speech. Experimental results on the L2-ARCTIC and VCTK datasets demonstrate that the proposed method outperforms existing retrieval-based and generative baselines in both objective and subjective evaluations, achieving high intelligibility, speaker similarity, and naturalness. Demo page is available at https://aa-knn-ac.github.io/demo/.

cs.SD↗
arXiv2026

ProCraft: Instruction-Driven Music Production from Free-Form Text

Music production typically involves multiple editing operations that each require specialized tools and expertise. Existing text-guided music editing systems support few operations or instruments, operate on short clips, or handle only one operation per instruction, unlike real workflows where several changes are requested at once. Here, we present ProCraft, a lightweight 0.5B-parameter autoregressive codec language model that edits 30s music excerpts end-to-end from free-form instructions, including compound requests. ProCraft supports seven operations across eight instruments, from tasks with a unique target (e.g., track removal) to creative ones (e.g., accompaniment generation). We align intermediate representations with pretrained music embeddings and pitch and instrument targets to preserve content outside the edit, and gated cross-attention in every decoder layer to keep generation tied to the instruction. To train on compound requests, we introduce On-the-fly Editing-pair Synthesis, which generates paired instructions and audio combining track-level operations with effect changes. ProCraft performs competitively against task-specific systems on single-task editing, and outperforms four free-text editors on compound instructions for track removal, extraction and denoising on real (MoisesDB) and synthetic (Slakh) recordings. In a user study (16 listeners), ProCraft was rated highest on removal and denoising and on-par with a 16-times larger model on extraction and instrument transfer. Ablations further show that representation alignment and cross-attention reduce leakage from unwanted tracks and raise editing success rates, especially on out-of-domain data.

cs.SD↗