UCSC-VLAA/story-iter
A training-free iterative diffusion model framework for generating coherent image sequences from long narrative text.
Not currently ranked — collecting fresh signals.
star history
Story-Iter is a research implementation for long story visualization using diffusion models. It introduces an external iterative paradigm that refines each generated image by incorporating reference images from previous rounds, addressing semantic consistency and fine-grained interaction challenges in multi-frame visual storytelling. The framework proposes a plug-and-play global reference cross-attention mechanism that operates without additional training.
Frequently asked
- What is UCSC-VLAA/story-iter?
- A training-free iterative diffusion model framework for generating coherent image sequences from long narrative text.
- Is story-iter open source?
- Yes — UCSC-VLAA/story-iter is open source, released under the MIT license.
- What language is story-iter written in?
- UCSC-VLAA/story-iter is primarily written in Python.
- How popular is story-iter?
- UCSC-VLAA/story-iter has 959 stars on GitHub.
- Where can I find story-iter?
- UCSC-VLAA/story-iter is on GitHub at https://github.com/UCSC-VLAA/story-iter.