Shypd.aiThe largest open index of AI toolsSEED-Story is a Content & Design tool that generates multimodal long stories with coherent narrative texts and consistent images. It is powered by a Multimodal Large Language Model (MLLM) and includes the StoryStream dataset for training and benchmarking.
Trending
About
SEED-Story is an advanced Multimodal Large Language Model (MLLM) developed by TencentARC, designed for generating comprehensive and coherent long stories. This tool excels at creating narrative texts alongside images that maintain character and style consistency throughout the story. It can generate stories spanning up to 25 multimodal sequences, even when trained on fewer. A key feature is its ability to produce diverse stories from the same initial image but different opening texts, allowing for varied narrative paths. SEED-Story utilizes a three-stage method involving an SD-XL-based de-tokenizer, an MLLM for next-word prediction and image feature regression, and fine-tuning of SD-XL for enhanced consistency. It also introduces StoryStream, a large-scale dataset for training and benchmarking multimodal story generation.
Capabilities
Pricing & Plans
Open Source
Free
FAQs
Trending
