# SoulX-FlashTalk: Real-Time Infinite Streaming of Audio-Driven Avatars via Self-Correcting Bidirectional Distillation

arXiv: https://arxiv.org/abs/2512.23379 · HTML: https://mlsanity.com/paper/2512.23379/

## Key facts

- Cited by 10 papers in the map (data through 14 Oct 2026).
- Placed in Audio, speech and music, one of 14 directions in Diffusion models.
- Builds on 24 of its 28 references that are in the map.
- First posted on arXiv 29 Dec 2025.

## Placed in

- Diffusion models / [Audio, speech and music](https://mlsanity.com/diffusion/audio-speech-and-music/)

## Builds on (most cited first, 24 of 24)

- [Learning Transferable Visual Models From Natural Language Supervision](https://mlsanity.com/paper/2103.00020/) (2021)
- [Exploring the Limits of Transfer Learning with a Unified Text-to-Text Transformer](https://mlsanity.com/paper/1910.10683/) (2019)
- [Scalable Diffusion Models with Transformers](https://mlsanity.com/paper/2212.09748/) (2022)
- [Classifier-Free Diffusion Guidance](https://mlsanity.com/paper/2207.12598/) (2022)
- [wav2vec 2.0: A Framework for Self-Supervised Learning of Speech Representations](https://mlsanity.com/paper/2006.11477/) (2020)
- [Wan: Open and Advanced Large-Scale Video Generative Models](https://mlsanity.com/paper/2503.20314/) (2025)
- [VBench: Comprehensive Benchmark Suite for Video Generative Models](https://mlsanity.com/paper/2311.17982/) (2023)
- [One-step Diffusion with Distribution Matching Distillation](https://mlsanity.com/paper/2311.18828/) (2023)
- [PyTorch FSDP: Experiences on Scaling Fully Sharded Data Parallel](https://mlsanity.com/paper/2304.11277/) (2023)
- [Self Forcing: Bridging the Train-Test Gap in Autoregressive Video Diffusion](https://mlsanity.com/paper/2506.08009/) (2025)
- [FlashAttention-3: Fast and Accurate Attention with Asynchrony and Low-precision](https://mlsanity.com/paper/2407.08608/) (2024)
- [Q-Align: Teaching LMMs for Visual Scoring via Discrete Text-Defined Levels](https://mlsanity.com/paper/2312.17090/) (2023)
- [Ring Attention with Blockwise Transformers for Near-Infinite Context](https://mlsanity.com/paper/2310.01889/) (2023)
- [DeepSpeed Ulysses: System Optimizations for Enabling Training of Extreme Long Sequence Transformer Models](https://mlsanity.com/paper/2309.14509/) (2023)
- [LivePortrait: Efficient Portrait Animation with Stitching and Retargeting Control](https://mlsanity.com/paper/2407.03168/) (2024)
- [Self-Forcing++: Towards Minute-Scale High-Quality Video Generation](https://mlsanity.com/paper/2510.02283/) (2025)
- [OmniAvatar: Efficient Audio-Driven Avatar Video Generation with Adaptive Body Animation](https://mlsanity.com/paper/2506.18866/) (2025)
- [Wan-S2V: Audio-Driven Cinematic Video Generation](https://mlsanity.com/paper/2508.18621/) (2025)
- [StableAvatar: Infinite-Length Audio-Driven Avatar Video Generation](https://mlsanity.com/paper/2508.08248/) (2025)
- [Ditto: Motion-Space Diffusion for Controllable Realtime Talking Head Synthesis](https://mlsanity.com/paper/2411.19509/) (2024)
- [InfiniteTalk: Audio-driven Video Generation for Sparse-Frame Video Dubbing](https://mlsanity.com/paper/2508.14033/) (2025)
- [SpeakerVid-5M: A Large-Scale High-Quality Dataset for Audio-Visual Dyadic Interactive Human Generation](https://mlsanity.com/paper/2507.09862/) (2025)
- [EchoMimicV3: 1.3B Parameters are All You Need for Unified Multi-Modal and Multi-Task Human Animation](https://mlsanity.com/paper/2507.03905/) (2025)
- [AnyTalker: Scaling Multi-Person Talking Video Generation with Interactivity Refinement](https://mlsanity.com/paper/2511.23475/) (2025)

## Built on by (most cited first, 10 of 10)

- [LongCat-Video-Avatar 1.5 Technical Report](https://mlsanity.com/paper/2605.26486/) (2026)
- [Video Generation Models as World Models: Efficient Paradigms, Architectures and Algorithms](https://mlsanity.com/paper/2603.28489/) (2026)
- [Vidu S2: Real-Time Interactive, Editable, and Spatial Video Generation](https://mlsanity.com/paper/2609.11638/) (2026)
- [StreamChar: Long-Horizon Streaming Character Audio-Video Generation with Decoupled Orchestration](https://mlsanity.com/paper/2605.25659/) (2026)
- [TurboTalk: Progressive Distillation for One-Step Audio-Driven Talking Avatar Generation](https://mlsanity.com/paper/2604.14580/) (2026)
- [Avatar-Forever: Decoupled Parallel Training for High-Quality Real-Time Infinite Avatars](https://mlsanity.com/paper/2608.12107/) (2026)
- [DynaForcing: Overcoming Dynamic Collapse in Self-Forcing Distillation for Streaming Avatar Generation](https://mlsanity.com/paper/2608.17707/) (2026)
- [EMOSH: Expressive Motion and Shape Disentanglement for Human Animation](https://mlsanity.com/paper/2606.28026/) (2026)
- [Vorch-Streamer: Extending Human Audio-Visual Generation to Real-Time Long-Form Streaming](https://mlsanity.com/paper/2608.05663/) (2026)
- [SWT: Self-Supervised Video Object Segmentation via Sliding, Wavelet and Transportation](https://mlsanity.com/paper/2609.31725/) (2026)

## Abstract (arXiv)

Deploying massive diffusion models for real-time, infinite-duration, audio-driven avatar generation presents a significant engineering challenge, primarily due to the conflict between computational load and strict latency constraints. Existing approaches often compromise visual fidelity by enforcing strictly unidirectional attention mechanisms or reducing model capacity. To address this problem, we introduce SoulX-FlashTalk, a 14B-parameter framework optimized for high-fidelity real-time streaming. Diverging from conventional unidirectional paradigms, we use a Self-correcting Bidirectional Distillation strategy that retains bidirectional attention within video chunks. This design preserves critical spatiotemporal correlations, significantly enhancing motion coherence and visual detail. To ensure stability during infinite generation, we incorporate a Multi-step Retrospective Self-Correction Mechanism, enabling the model to autonomously recover from accumulated errors and preventing collapse. Furthermore, we engineered a full-stack inference acceleration suite incorporating hybrid sequence parallelism, Parallel VAE, and kernel-level optimizations. Extensive evaluations confirm that SoulX-FlashTalk is the first 14B-scale system to achieve a sub-second start-up latency (0.87s) while reaching a real-time throughput of 32 FPS, setting a new standard for high-fidelity interactive digital human synthesis.