# FantasyTalking2: Timestep-Layer Adaptive Preference Optimization for Audio-Driven Portrait Animation

arXiv: https://arxiv.org/abs/2508.11255 · HTML: https://mlsanity.com/paper/2508.11255/

## Key facts

- Cited by 3 papers in the map (data through 14 Oct 2026).
- Placed in Diffusion and multimodal model alignment, one of 10 directions in LLM alignment / RLHF / preference optimisation.
- Builds on 42 of its 60 references that are in the map.
- First posted on arXiv 15 Aug 2025.

## Placed in

- LLM alignment / RLHF / preference optimisation / [Diffusion and multimodal model alignment](https://mlsanity.com/alignment/diffusion-and-multimodal-model-alignment/)

## Builds on (most cited first, 30 of 42)

- [Training language models to follow instructions with human feedback](https://mlsanity.com/paper/2203.02155/) (2022)
- [The Llama 3 Herd of Models](https://mlsanity.com/paper/2407.21783/) (2024)
- [Direct Preference Optimization: Your Language Model is Secretly a Reward Model](https://mlsanity.com/paper/2305.18290/) (2023)
- [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)
- [Learning to summarize from human feedback](https://mlsanity.com/paper/2009.01325/) (2020)
- [RAFT: Recurrent All-Pairs Field Transforms for Optical Flow](https://mlsanity.com/paper/2003.12039/) (2020)
- [HunyuanVideo: A Systematic Framework For Large Video Generative Models](https://mlsanity.com/paper/2412.03603/) (2024)
- [SimPO: Simple Preference Optimization with a Reference-Free Reward](https://mlsanity.com/paper/2405.14734/) (2024)
- [A General Theoretical Paradigm to Understand Learning from Human Preferences](https://mlsanity.com/paper/2310.12036/) (2023)
- [Animate Anyone: Consistent and Controllable Image-to-Video Synthesis for Character Animation](https://mlsanity.com/paper/2311.17117/) (2023)
- [MediaPipe: A Framework for Building Perception Pipelines](https://mlsanity.com/paper/1906.08172/) (2019)
- [Q-Align: Teaching LMMs for Visual Scoring via Discrete Text-Defined Levels](https://mlsanity.com/paper/2312.17090/) (2023)
- [SadTalker: Learning Realistic 3D Motion Coefficients for Stylized Audio-Driven Single Image Talking Face Animation](https://mlsanity.com/paper/2211.12194/) (2022)
- [Improving Video Generation with Human Feedback](https://mlsanity.com/paper/2501.13918/) (2025)
- [Champ: Controllable and Consistent Human Image Animation with 3D Parametric Guidance](https://mlsanity.com/paper/2403.14781/) (2024)
- [3D Morphable Face Models -- Past, Present and Future](https://mlsanity.com/paper/1909.01815/) (2019)
- [LivePortrait: Efficient Portrait Animation with Stitching and Retargeting Control](https://mlsanity.com/paper/2407.03168/) (2024)
- [VideoScore: Building Automatic Metrics to Simulate Fine-grained Human Feedback for Video Generation](https://mlsanity.com/paper/2406.15252/) (2024)
- [EchoMimic: Lifelike Audio-Driven Portrait Animations through Editable Landmark Conditions](https://mlsanity.com/paper/2407.08136/) (2024)
- [AniPortrait: Audio-Driven Synthesis of Photorealistic Portrait Animation](https://mlsanity.com/paper/2403.17694/) (2024)
- [Identity-Preserving Text-to-Video Generation by Frequency Decomposition](https://mlsanity.com/paper/2411.17440/) (2024)
- [VisionReward: Fine-Grained Multi-Dimensional Human Preference Learning for Image and Video Generation](https://mlsanity.com/paper/2412.21059/) (2024)
- [DiffTalk: Crafting Diffusion Models for Generalized Audio-Driven Portraits Animation](https://mlsanity.com/paper/2301.03786/) (2023)
- [OmniHuman-1: Rethinking the Scaling-Up of One-Stage Conditioned Human Animation Models](https://mlsanity.com/paper/2502.01061/) (2025)
- [Phantom: Subject-consistent video generation via cross-modal alignment](https://mlsanity.com/paper/2502.11079/) (2025)
- [Beyond One-Preference-Fits-All Alignment: Multi-Objective Direct Preference Optimization](https://mlsanity.com/paper/2310.03708/) (2023)
- [VideoDPO: Omni-Preference Alignment for Video Diffusion Generation](https://mlsanity.com/paper/2412.14167/) (2024)
- [Deep Reinforcement Learning for Multi-objective Optimization](https://mlsanity.com/paper/1906.02386/) (2019)
- [OmniAvatar: Efficient Audio-Driven Avatar Video Generation with Adaptive Body Animation](https://mlsanity.com/paper/2506.18866/) (2025)

## Built on by (most cited first, 3 of 3)

- [From Inpainting to Editing: Unlocking Robust Mask-Free Visual Dubbing via Generative Bootstrapping](https://mlsanity.com/paper/2512.25066/) (2025)
- [VividAnimator: An End-to-End Audio and Pose-driven Half-Body Human Animation Framework](https://mlsanity.com/paper/2510.10269/) (2025)
- [JoyStreamer: Unlocking Highly Expressive Avatars via Harmonized Text-Audio Conditioning](https://mlsanity.com/paper/2602.00702/) (2026)

## Abstract (arXiv)

Recent advances in audio-driven portrait animation have demonstrated impressive capabilities. However, existing methods struggle to align with fine-grained human preferences across multiple dimensions, such as motion naturalness, lip-sync accuracy, and visual quality. This is due to the difficulty of optimizing among competing preference objectives, which often conflict with one another, and the scarcity of large-scale, high-quality datasets with multidimensional preference annotations. To address these, we first introduce Talking-Critic, a multimodal reward model that learns human-aligned reward functions to quantify how well generated videos satisfy multidimensional expectations. Leveraging this model, we curate Talking-NSQ, a large-scale multidimensional human preference dataset containing 410K preference pairs. Finally, we propose Timestep-Layer adaptive multi-expert Preference Optimization (TLPO), a novel framework for aligning diffusion-based portrait animation models with fine-grained, multidimensional preferences. TLPO decouples preferences into specialized expert modules, which are then fused across timesteps and network layers, enabling comprehensive, fine-grained enhancement across all dimensions without mutual interference. Experiments demonstrate that Talking-Critic significantly outperforms existing methods in aligning with human preference ratings. Meanwhile, TLPO achieves substantial improvements over baseline models in lip-sync accuracy, motion naturalness, and visual quality, exhibiting superior performance in both qualitative and quantitative evaluations. Ours project page: https://fantasy-amap.github.io/fantasy-talking2/