README.md

---
license: cc0-1.0
task_categories:
- text-to-image
- image-to-text
language:
- en
size_categories:
- 10K<n<100K
---

This dataset features a collection of high-quality images sourced from Pexels and captioned using the Gemini-1.5-Flash API. This dataset is designed to provide accurate, detailed descriptions of various visual content, suitable for text-to-image tasks, training AI models, and more.





Gemini promt:
"Describe this image, for a text-to-image train to be accurate, max 74 tokens. (the common theme between these images is '{theme}'), prefer the use of ',' dont use '.' and there is no need to have a point at the end, avoid double spaces, don't forget to label the image with one indication of media examples (count this as 1 token): graphite_(artwork), pencil_(artwork), digital_media_(artwork), digital_media_(photo), digital_media_(photography), digital_media_(illustration), digital_media_(3D_artwork), digital_media_(vector_art), digital_media_(motion_graphics), digital_media_(pixel_art), digital_media_(digital_collage), digital_media_(landscape_photo), digital_media_(portrait_photo), digital_media_(macro_photo), digital_media_(street_photo), digital_media_(aerial_photo), digital_media_(concept_art), digital_media_(character_design), digital_media_(environment_art), digital_media_(digital_painting), digital_media_(web_design), digital_media_(graphic_design), digital_media_(UI_design), digital_media_(game_art), traditional_media_(artwork), sculpture_(artwork)."