Datasets:

agentlans
/

FLUX.1-schnell-random

license: cc-by-4.0
task_categories:
  - text-to-image
tags:
  - flux
  - flux-schnell
  - latent-space
  - synthetic
size_categories:
  - 1K<n<10K
language:
  - en
  - de

Flux 1 Schnell Random Images

A dataset of 3000 synthetic images (1024 x 1024) generated to explore the latent space of black-forest-labs/FLUX.1-schnell in a completely randomized, unbiased way.

Generation Parameters

Model Checkpoint: flux1-schnell-fp8.safetensors
Sampler: euler_ancestral | Steps: 4 | CFG: 1.0 | Flux Guidance: 3.5
Seed: Random

Prompt Configuration

Prompt: Dynamically assembled string using up to the maximum limit of 512 random tokens.

Example Row

{
  "id": "ff47550d-904b-43e1-a4b8-925c3b2b3734",
  "image": "<binary blob>",
  "prompt": "duplicate Unfortunatelyopathygoing bouton spinalDetailedutăkar Canon folosindvatiguru Bureauvalent faci workers Beiträge présent180 Classroomouvri Auction documented Tokyo vivant about furnished Alpine Stufe definitive Companies hinausKollekt Yefertig scanning Comedy contient precursorProbably transactionsarii euch pass Herz telecom deserves Tele started vest railway retourproduced Vier deosebit beads array ZhaoSYSAvec cutie studied Request Inhaltissue dwelling census Vanessaconciliation..."
}

Limitations

Fixed Resolution: All generated images are strictly set to 1024 x 1024 resolution.
Text Artifacts: Nonsensical random token prompts result in garbled, nonsensical text rendered directly inside the images.
Visual Glitches: Strange visual artifacts occur (e.g., headless people, floating cups, walls dissolving into the sky).
Model Biases: Inherent latent biases of the Flux model surface frequently (e.g., disproportionately high occurrences of pineapples, turtles, butterflies, and Starbucks logos).

License

This dataset is released under the CC-BY-4.0 license.