Research map / Image generation / Editing and style transfer
IdentityStory: Taming Your Identity-Preserving Generator for Human-Centric Story Generation
Donghao Zhou, Jingyu Lin, Guibao Shen, Quande Liu, Jialin Gao, Lihao Liu, Lan Du, Cunjian Chen, Chi-Wing Fu, Xiaowei Hu, Pheng-Ann Heng
Read the paperCheck your idea against 137,217 papers
AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Measured effects reported in the paper
| What was changed | Where | Measured | Effect |
|---|---|---|---|
| replace attention-sharing and prompt-based consistency with identity-preserving generator plus identity extraction | IdentityStory · ConsiStory-Human | Face-Sim (%): 27.1 -> 55.5 vs StoryDiffusion | Improved |
| add Iterative Identity Discovery (SVD filtering) instead of naive averaging of identity embeddings (ablation) | IdentityStory · ConsiStory-Human | character consistency / Face-Sim | Improved |
| increase iteration number of Iterative Identity Discovery (ablation) | IdentityStory · 20 character descriptions | Face-Sim | Improved |
| add Re-denoising Identity Injection instead of direct generation with the identity-preserving generator (ablation) | IdentityStory · ConsiStory-Human | text alignment | Improved |
| use layout-guided injection instead of averaging cached noisy images with identity generator output (ablation) | IdentityStory · ConsiStory-Human | visual quality | Improved |
| use progressive mask dilation instead of fixed masks during re-denoising (ablation) | IdentityStory · ConsiStory-Human | artifacts | Improved |
| start re-denoising at timestep t'=40 (ablation) | IdentityStory · ConsiStory-Human | image harmony vs identity fidelity | Improved |
| IdentityStory versus ConsiStory on CLIP-T | IdentityStory · ConsiStory-Human | CLIP-T (%): 35.5 -> 35.4 | Hurt |
| use IdentityStory instead of 1Prompt1Story in GPT-4o judge | IdentityStory · ConsiStory-Human (GPT-4o judge) | character consistency score: 78.31 -> 82.83 | Improved |
Closest papers by meaning
- One-Prompt-One-Story: Free-Lunch Consistent Text-to-Image Generation Using a Single Prompt (ICLR 2025 · similarity 0.88)
- Storynizor: Consistent Story Generation via Inter-Frame Synchronized and Shuffled ID Injection (AAAI 2025 · similarity 0.88)
- Storybooth: Training-Free Multi-Subject Consistency for Improved Visual Storytelling (ICLR 2025 · similarity 0.86)
- Advancing Narrative Long Video Generation via Training-Free Identity-Aware Memory (NeurIPS 2026 · similarity 0.86)
- DreamSalon: A Staged Diffusion Framework for Preserving Identity-Context in Editable Face Generation (CVPR 2024 · similarity 0.86)
- Intelligent Grimm - Open-ended Visual Storytelling via Latent Diffusion Models (CVPR 2024 · similarity 0.86)
Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.
