NARA

NARA Image · bucket-backed studio

Generate boldly.
Keep everything.

Eight pro image tools — Flux, Imagen, SDXL and friends behind one prompt box — and every output lands in a gallery persisted to S3. A studio, not a slot machine.

Open the studio50 free credits
before · rawafter · nara
hover to process
flux · imagen · sdxl · ideogramscroll
Text to imageUpscaleRetouchRemove BGBackgroundGenerative fillResizeTransparent

/manifesto

/tools

Explore NARA Image tools

Eight operations, one credit meter. Each tool reads and writes the same gallery — outputs chain into the next edit.

prompt → pixels

Text to image

Flux, Imagen, SDXL and friends. Type once, render four variants.

open

×4 detail

Upscale

Push renders to 4K-ready sharpness without the plastic AI sheen.

open

fix, don't redo

Retouch

Mask a region, describe the fix, keep every other pixel untouched.

open

clean cutouts

Remove background

Hair-level mattes in one pass — alpha channel included, no halo.

open

new scene

Replace background

Drop the subject into a studio sweep, a product set, anywhere.

open

in + outpaint

Generative fill

Grow the canvas or heal a hole — the seams blend themselves.

open

any aspect

Smart resize

Reframe to 1:1, 16:9, 4:5 — content-aware, never a dumb crop.

open

straight to alpha

Transparent

Generate directly onto transparency — composite-ready PNGs.

open

/loop

One loop, four beats.

01Prompt

Type it like you mean it.

Pick a frontier model — Flux 1.1 Pro, Imagen 4, SDXL Lightning — set size and seed, and write. The cost is shown before you commit a single credit.

image.naravirtual.ai/tools/text-to-image
flux-1.1-proimagen-4sdxl-lightning
a lighthouse at blue hour, long exposure, cinematic fog▍
2048 × 2048 · n=4 · seed autorender — 4 cr

02Generate

Four variants, one hold.

Credits are reserved, not burned: the studio holds the estimate, renders your variants in seconds, and only commits what you keep.

image.naravirtual.ai/studio
4 variants · 3.1sheld 4 cr · commit on save

03Refine

Edit the pixels, not the prompt.

Upscale ×4, retouch a masked region, strip the background — every operation reads the last output and writes a new version. Nothing is re-rolled from scratch.

nara — refine

$ upscale --scale 4

→ 4096 × 4096 · detail preserved · 2 cr

$ retouch --mask sky --prompt "denser fog"

→ region re-rendered · rest untouched · 3 cr

$ remove-bg

→ alpha matte · hair-level edges · 1 cr

$ ▍

04Ship

It's in your bucket, forever.

Every version lands in your gallery, persisted to S3 with its prompt, seed, and lineage. Come back next month and keep editing where you left off.

image.naravirtual.ai/gallery
s3://nara-images/u_2214/render_8841.png
persisted · versioned · yours

/models

Best-in-class models, one meter.

0

studio tools

0+

frontier models

0%

outputs persisted

0

free credits to start

flux-1.1-proflux-1.1-devsdxl-lightningimagen-4ideogram-v3gpt-image-1stable-image-ultra

/gallery

studies · rendered with the tools above

flux-1.1-pro · seed 48122
imagen-4 · seed 90417
ideogram-v3 · seed 22903
sdxl-lightning · seed 5511
flux-1.1-dev · seed 77260
stable-image-ultra · seed 31038

50 credits on sign-up. No card. Your first render is about ninety seconds away.

Open the studio

text-to-image · upscale · retouch · remove-bg