CoolFace
Datasetpublic

retowyss/Persona-Fluxed-10k-2608

Persona Fluxed 10k Synthetic persona portraits rendered with FLUX.2-klein-4b (8-step, 1024x1024) from the NVIDIA Nemotron-Personas-* datasets. Each persona is grounded in real-world demographic, geographic and personality-trait distributions for its country (CC BY 4.0 source; no real people). Currently Nemotron-Personas exist for: USA — English Japan — Japanese India — English, Hindi Brazil — Portuguese Singapore — English France — French Korea — Korean El Salvador — Spanish… See the full description on the dataset page: https://huggingface.co/datasets/retowyss/Persona-Fluxed-10k-2608.

sourceHugging Facecc-by-4.0updated 1mo agoView on Hugging Face
0likes400downloads
Dataset Card

Persona Fluxed 10k

Synthetic persona portraits rendered with FLUX.2-klein-4b (8-step, 1024x1024) from the NVIDIA Nemotron-Personas-* datasets. Each persona is grounded in real-world demographic, geographic and personality-trait distributions for its country (CC BY 4.0 source; no real people).

Currently Nemotron-Personas exist for:

  • —USA — English
  • —Japan — Japanese
  • —India — English, Hindi
  • —Brazil — Portuguese
  • —Singapore — English
  • —France — French
  • —Korea — Korean
  • —El Salvador — Spanish
  • —Vietnam — Vietnamese
  • —Belgium — Dutch, French, German, English

Structure

One subset per country (10), each with a single train split: PNG files + a metadata.csv in the same directory.

Columns

columnmeaning
file_nameimage file in the same directory
uuidsource persona id (stable across the 6 fields of one person)
fieldpersona facet: professional / sports / arts / travel / culinary / plain
persona_textraw source description (native language)
promptexact prompt sent to the model ("A photo of: " + persona_text)
styleVLM-classified visual style: Photo / Photo-Collage / Digital Art / Other
countryconfig slug: usa, japan, india, brazil, singapore, france, korea, el-salvador, vietnam, belgium
source_datasetoriginating nvidia/Nemotron-Personas-* repo id
country_metasource's native country label
age, sex, marital_status, education_level, occupationdemographic attributes (native-language values, empty where absent)

Notes

  • —~10k portraits (10 countries x 167 personas x 6 fields).
  • —travel personas frequently render as multi-panel collages; arts personas drift toward stylized art most often. Filter on style as needed.
  • —Same uuid appears up to 6 times (once per facet) — join on uuid to group the facets of one person.

License

Both Nemotron and FLUX.2-klein-4b have permissive licenses.

  • —The docs and metadata of this dataset shall be considered CC BY 4.0
  • —No claims or guarantees are made in regards to images - provided as is.