Research map / Vision-language and multimodal / Human and embodied modeling
OmniLottie: Generating Vector Animations via Parameterized Lottie Tokens
Yiying Yang, Wei Cheng, Sijin Chen, Honghao Fu, Xianfang Zeng, Yujun Cai, Gang Yu, Xingjun Ma
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 |
|---|---|---|---|
| Add Lottie tokenizer instead of fine-tuning on raw JSON (ablation) | Qwen2.5-VL · Text-to-Lottie | success rate: 13.4% -> 97.3% | Improved |
| Add Lottie tokenizer instead of fine-tuning on raw JSON (ablation) | Qwen2.5-VL · Text-to-Lottie | FVD: 459.39 -> 269.50 | Improved |
| Add Lottie tokenizer instead of fine-tuning on raw JSON (ablation) | Qwen2.5-VL · Text-Image-to-Lottie | FVD: 476.23 -> 359.56 | Improved |
| Add Lottie tokenizer instead of fine-tuning on raw JSON (ablation) | Qwen2.5-VL · Video-to-Lottie | success rate: 10.2% -> 90.7% | Improved |
| Add Lottie tokenizer instead of fine-tuning on raw JSON (ablation) | Qwen2.5-VL · Video-to-Lottie | PSNR: 17.23 -> 19.05 | Improved |
| Fine-tune pretrained Qwen2.5-VL (any form) vs zero-shot prompting (ablation) | Qwen2.5-VL 3B · Text-to-Lottie | success rate: 0.0% -> 13.4% (raw JSON) | Improved |
| Use OmniLottie instead of GPT-5 | Lottie generation · Text-to-Lottie (Real) | FVD: 715.73 -> 202.14 | Improved |
| Use OmniLottie instead of Recraft | Text-to-Lottie · Real subset | object alignment: 4.70 vs 4.44 | Mixed |
| Use OmniLottie instead of Recraft | Text-to-Lottie · Real subset | motion alignment: 4.68 -> 5.94 | Improved |
| Use OmniLottie instead of optimization-based AniClipart/Livesketch | Text-Image-to-Lottie · Real subset | generation time (s): 1212.34 -> 88.57 | Improved |
| Move from Real to Synthetic subset | OmniLottie · Text-to-Lottie | success rate: 88.3% -> 82.1% | Hurt |
Closest papers by meaning
- Animation2Code: Evaluating Temporal Visual Reasoning in Video-to-Code Generation (NeurIPS 2026 · similarity 0.84)
- VecGlypher: Unified Vector Glyph Generation with Language Models (CVPR 2026 · similarity 0.83)
- How to Move Your Dragon: Text-to-Motion Synthesis for Large-Vocabulary Objects (ICML 2025 · similarity 0.83)
- MotiF: Making Text Count in Image Animation with Motion Focal Loss (CVPR 2025 · similarity 0.83)
- OmniMotionGPT: Animal Motion Generation with Limited Data (CVPR 2024 · similarity 0.83)
- Chat2SVG: Vector Graphics Generation with Large Language Models and Image Diffusion Models (CVPR 2025 · similarity 0.83)
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.
