Tensor.Art
Create

[SERIES] A mosaic of nations via vanilla 'Stable Diffusion 3.5 Large'

Updated:

I am proudly presenting my series of portraits (head-and-shoulders studio shots) that showcase a wide range of nations across the globe. Each portrait is carefully crafted to reflect the distinct facial features of the given ethnic group in its most natural form.

The name is A mosaic of nations via vanilla Stable Diffusion 3.5 Large.

This rich series aims to present at least 10 portraits (head-and-shoulders studio shots) per group, and it’s all about…

[a] Diversity

FIFTY💥 ethnic groups (A to Z), along with the Ethnically undefined group, together represent over 75% of the world’s population:

  1. _Ethnically undefined

  2. Amhara

  3. Armenian (ethnicity rather than nationality)

  4. Ashkenazi Jewish

  5. Baltic

  6. Bengali

  7. Berber/Amazigh

  8. biracial Black-American

  9. biracial Korean-American

  10. Black-American

  11. Caribbean

  12. Caucasus native (Eurasian highlander)

  13. Celtic

  14. Dinka/Nuer

  15. Dravidian

  16. Eastern Slavic

  17. Euro-Mediterranean

  18. Fulani

  19. Germanic

  20. Han

  21. Himalaya native (Nepalese highlander)

  22. Hispanic SouthAmerican

  23. Iberian

  24. Kalaallit (Greenlandic Inuit)

  25. Khaleeji/Gulfian

  26. Kurdish

  27. Levantine-Arab

  28. Maghrebi-Arab

  29. Magyar

  30. Miao

  31. Navajo/Diné

  32. North American

  33. Pardo (biracial Brazilian)

  34. Pashtun

  35. Persian

  36. Polynesian

  37. Punjabi

  38. Quechua

  39. Sámi

  40. Scandinavian

  41. Sephardi Jewish

  42. Somali (ethnicity rather than nationality)

  43. Southeast Asian

  44. Southern Slavic

  45. Tatar

  46. Uzbek (ethnicity rather than nationality)

  47. Western European

  48. Western Slavic

  49. Yakut/Sakha

  50. Yamato

  51. Zhuang

[b] Purity

⬩ Vanilla model – Untuned and unmerged for optimal performance.
⬩ No theming, stylization, or re-texturing – LoRA not applied; should be done afterward at your discretion, via inpainting (essentially the "LoRA" process).
⬩ No post-processing effects – No smoothing, glow, HDR, etc., simulating the real world.
⬩ No personal enhancements – No nonessential apparel (headwear, neckwear, palmwear, etc.), wearable adornments (eyewear, jewelry, wristwear, etc.), or beauty addons (makeup, hair dye, false eyelashes, etc.), preserving a 99.88% natural subject.
⬩ No upscaling – Upscale set to OFF; should be done afterward at your discretion.
⬩ No refinement – ADetailer set to OFF; should be done afterward at your discretion, via inpainting, face-swap, or bg-swap.
⬩ No excessive or redundant prompting – Concise input for optimal adherence.

[c] Quality

⬩ Fine detail and smooth gradient – Max (60) sampling steps, approximately twice the average.
⬩ Generous pixel count – 1.05 megapixels, equivalent to 1024×1024—more than enough for most (online) cases. Upscale should be done afterward at your discretion.
⬩ Crisp footage – No compression artifacts, unintended blur, or subject distortion. In other words, "The lake is so lucid, you can see the bottom."

[d] Professionality

⬩ Studio conditions with controlled lighting – Just the right amount of contrast, brightness, saturation, and sharpness.
⬩ True-to-life visuals – They won’t know it’s AI-generated.
⬩ Optimal aspect ratio – 16:11, closely aligned with the popular 2:3 ratio. Perfect for portraits, good for a wide variety of compositions.
⬩ Editor-friendly background – Plain teal surface for effortless background swap/edit.

Just a few examples

[SERIES] A mosaic of nations via vanilla Stable Diffusion 3.5 Large: Zhuang

[SERIES] A mosaic of nations via vanilla Stable Diffusion 3.5 Large: Southeast Asian

[SERIES] A mosaic of nations via vanilla Stable Diffusion 3.5 Large: Punjabi

Discover more

8
0