# DynaForcing: Overcoming Dynamic Collapse in Self-Forcing Distillation for Streaming Avatar Generation

arXiv: https://arxiv.org/abs/2608.17707 · HTML: https://mlsanity.com/paper/2608.17707/

## Key facts

- Cited by 1 papers in the map (data through 14 Oct 2026).
- Builds on 24 of its 44 references that are in the map.
- First posted on arXiv 18 Aug 2026.

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

- [DINOv2: Learning Robust Visual Features without Supervision](https://mlsanity.com/paper/2304.07193/) (2023)
- [VBench: Comprehensive Benchmark Suite for Video Generative Models](https://mlsanity.com/paper/2311.17982/) (2023)
- [CNN-generated images are surprisingly easy to spot... for now](https://mlsanity.com/paper/1912.11035/) (2019)
- [A Lip Sync Expert Is All You Need for Speech to Lip Generation In The Wild](https://mlsanity.com/paper/2008.10010/) (2020)
- [Self Forcing: Bridging the Train-Test Gap in Autoregressive Video Diffusion](https://mlsanity.com/paper/2506.08009/) (2025)
- [How far are we from solving the 2D & 3D Face Alignment problem? (and a dataset of 230,000 3D facial landmarks)](https://mlsanity.com/paper/1703.07332/) (2017)
- [Q-Align: Teaching LMMs for Visual Scoring via Discrete Text-Defined Levels](https://mlsanity.com/paper/2312.17090/) (2023)
- [Z-Image: An Efficient Image Generation Foundation Model with Single-Stream Diffusion Transformer](https://mlsanity.com/paper/2511.22699/) (2025)
- [LongLive: Real-time Interactive Long Video Generation](https://mlsanity.com/paper/2509.22622/) (2025)
- [Causal Forcing: Autoregressive Diffusion Distillation Done Right for High-Quality Real-Time Interactive Video Generation](https://mlsanity.com/paper/2602.02214/) (2026)
- [Looking to Listen at the Cocktail Party: A Speaker-Independent Audio-Visual Model for Speech Separation](https://mlsanity.com/paper/1804.03619/) (2018)
- [Hallo3: Highly Dynamic and Realistic Portrait Image Animation with Video Diffusion Transformer](https://mlsanity.com/paper/2412.00733/) (2024)
- [Reward Forcing: Efficient Streaming Video Generation with Rewarded Distribution Matching Distillation](https://mlsanity.com/paper/2512.04678/) (2025)
- [OmniAvatar: Efficient Audio-Driven Avatar Video Generation with Adaptive Body Animation](https://mlsanity.com/paper/2506.18866/) (2025)
- [EchoMimicV2: Towards Striking, Simplified, and Semi-Body Human Animation](https://mlsanity.com/paper/2411.10061/) (2024)
- [Wan-S2V: Audio-Driven Cinematic Video Generation](https://mlsanity.com/paper/2508.18621/) (2025)
- [Learning Few-Step Diffusion Models by Trajectory Distribution Matching](https://mlsanity.com/paper/2503.06674/) (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)
- [INFP: Audio-Driven Interactive Head Generation in Dyadic Conversations](https://mlsanity.com/paper/2412.04037/) (2024)
- [Diff-Instruct++: Training One-step Text-to-image Generator Model to Align with Human Preferences](https://mlsanity.com/paper/2410.18881/) (2024)
- [StreamAvatar: Streaming Diffusion Models for Real-Time Interactive Human Avatars](https://mlsanity.com/paper/2512.22065/) (2025)
- [Avatar Forcing: Real-Time Interactive Head Avatar Generation for Natural Conversation](https://mlsanity.com/paper/2601.00664/) (2026)
- [SoulX-FlashTalk: Real-Time Infinite Streaming of Audio-Driven Avatars via Self-Correcting Bidirectional Distillation](https://mlsanity.com/paper/2512.23379/) (2025)

## Built on by (most cited first, 1 of 1)

- [Where and When to Force: Routed Forcing for Streaming Avatars](https://mlsanity.com/paper/2609.30963/) (2026)

## Abstract (arXiv)

Audio-driven avatar generation requires realistic lip-sync, expressive motion, and real-time streaming. Recent work achieves the latter via self-forcing with Distribution Matching Distillation (DMD), but this paradigm suffers from a critical failure that has not been systematically characterized: dynamic collapse, where the student model converges to a near-static optimum with high perceptual quality but severely suppressed temporal dynamics. We trace this to two causes: the reverse KL objective in DMD, which biases toward low-motion modes, and unanchored self-conditioning, which creates a feedback loop that amplifies collapse. This is especially harmful for avatars, where even subtle motion loss breaks lip-sync and expression. To address this, we propose DynaForcing, a training framework with three complementary strategies applied at different levels. Specifically, Hybrid Forcing anchors rollouts to ground-truth dynamics at the data level to break the feedback loop. Dynamics-Aware Reward Regularization introduces explicit motion rewards via the RL interpretation of DMD to counteract the reverse KL bias at the loss level. Reference Perturbation perturbs reference images to decouple identity from static details, forcing the model to rely on audio for motion at the conditioning level. We further introduce computation graph pruning and gradient replay, reducing the GPU footprint of self-forcing by over an order of magnitude. Experiments show that DynaForcing recovers dynamics to teacher-comparable levels (Dyn-Deg: 0.31 -> 0.73, Sync-C: 7.03 -> 7.68) while improving visual quality, resolving the quality-dynamics trade-off throughout training without early stopping.