sjh9714/krea2-wildcards

Krea 2 Turbo Prompt and Output Catalog

This dataset contains 514 usable English prompts across 63 active categories. Every row keeps the exact prompt, the URL of its generated Krea 2 output, and generation provenance for filtering, retrieval, and prompt reuse.

It is the row-oriented export of the open Krea 2 Wildcards catalog. This is a community dataset, not an official Krea release.

Start here

• —Browse prompts beside their generated outputs.
• —Open the native six-node starter in Comfy Cloud.
• —Download the versioned v1.2.0 prompt pack and both ComfyUI workflows.

Fields

• —id, category, and title identify the catalog record.
• —prompt is the exact text used for the shown generation.
• —image_url points to the corresponding generated WebP output in the source repository.
• —params preserves generation provenance such as seed or hosted generation ID.
• —batch records the curation batch.
• —Editing records may also include source, strength, or attribution.

Use

from datasets import load_dataset

dataset = load_dataset("sjh9714/krea2-wildcards", split="train")
product_prompts = dataset.filter(lambda row: row["category"] == "product")
print(product_prompts[0]["prompt"])

Provenance and limitations

The images are AI-generated Krea 2 outputs and may contain typical generative-image artifacts. Prompts were retained only after their corresponding output was generated; the dataset does not add untested prompt text to inflate its size. The export contains no model weights and does not imply endorsement by Krea.

The prompt catalog, metadata, and repository assets are published under the source repository's MIT license. Krea states that compliant users own the outputs they generate; use of the Krea 2 model remains subject to the Krea 2 license and acceptable use policy.

Maintenance

prompts.json in the source repository is canonical. dataset/prompts.jsonl is generated with:

python3 scripts/build_dataset.py

Report data issues in the source repository.