Skip to content

SEED-Story

Generate long multimodal stories from images and text

Visit official site

What seed-story is

The model is described as a multimodal generator that builds long stories with consistent text and images.

Getting started

  • Clone the repository locally
  • Configure Python from requirements.txt
  • Download and prepare StoryStream

What to watch out for

  • Input images and text start the story
  • Chunking scripts aid training efficiency
  • Image style follows the narrative text

Pros & cons

✓ What we like

  • Image style and characters stay consistent
  • Tokenizer, De-Tokenizer and SEED-X weights are released

! What to watch out for

  • The repository must be cloned locally
  • The dataset needs preparation before training

FAQ

How many sequences are supported?

The material states up to 25, though training used up to 10.

What is released?

Pretrained Tokenizer, De-Tokenizer and the SEED-X base model are named.

Last reviewed: 2026-11-15

More AI image generation tools

View all →

How we review