Lumina
[ 00 ] IMAGE · TEXT · VIDEO · VOICE

From one prompt to production-ready in seconds.

Lumina is a generative studio and a single API for images, copy, video and voice. Models tuned to ship real assets into your product, not another gallery of cherry-picked demos.

Start generating, free Read the docs
$ pip install lumina  ·  40M+ generations served  ·  SOC 2
lumina · studio live
prompt › isometric glass robot, studio light, soft holographic rim 
lumina-image-v3 · 1024² 840ms

Shipping in production at

Helios Drift Studio Nimbus Pixelforge Atlas Media Verve Helios Drift Studio Nimbus Pixelforge Atlas Media Verve

[ 01 ] WATCH IT GENERATE

Type a prompt.
Watch the whole thing appear.

This isn't a screenshot. Scroll in and Lumina runs the prompt below for real, streaming an image set and the caption, token by token.

idle

Prompt

›  
lumina-image-v3 4 variations 1024²
Generated render seed · 1182
Generated render seed · 4027
Generated render seed · 7745
Generated render seed · 9310

Stream · lumina-text-v2

tokens

// product caption + alt-text, ready to ship

latency

·

tokens

·

cost

·

One /generate call returned the images, the caption and the alt-text above.

[ 02 ] FOUR MODALITIES, ONE BILL

Generate anything.
Through one interface.

Each model is built to be called in production: versioned, rate-limited, and deterministic when you need it to be.

img

Image.

Photoreal to vector, 4K, controllable seeds and reference images. Inpaint, outpaint and brand-lock a palette so every asset stays on-style.

Generated render
Generated render
Generated render
Generated render
txt

Text

copy, captions, alt-text. structured JSON out.

vid

Video

text-to-clip, 1080p, up to 12s with motion control.

Voice

Natural narration in 29 languages, clone-on-consent, SSML control.

One API, four models

Same auth, same SDK, streamed responses. Swap modality with one field.

840ms

[ 03 ] HOW IT WORKS

Prompt. Model. Output.

[ 01 ] prompt

Describe it

Plain words or a reference image. Pin a seed, lock a palette, set the aspect ratio. No prompt-engineering PhD required.

[ 02 ] model

We route it

Lumina picks the right model and resolution, runs it on warm GPUs, and streams partials back so nothing feels like waiting.

[ 03 ] output

Ship it

Get a hosted URL or raw bytes, plus metadata and a stable seed so you can reproduce the exact asset later.

0

Generations run

0

Production models

0

Median image latency

0

Voice languages

[ 04 ] PRICING

Pay for what you generate.

Every plan has all four models. Free credits to start, no card. Usage is metered per generation, billed monthly.

Free

For trying it on a side project.

€0/mo

  • ✓ 200 generations / month
  • ✓ All four models
  • ✓ 1024² images, 720p video
  • ✓ Community support
Start free

Pro

Most popular

For builders shipping real product.

€39/mo

  • ✓ 10,000 generations included
  • ✓ 4K images, 1080p · 12s video
  • ✓ Seeds, palettes & brand-lock
  • ✓ Streaming API + webhooks
  • ✓ Priority queue
Start 14-day trial

Scale

For teams generating at volume.

Let's talk

  • ✓ Volume pricing per generation
  • ✓ Dedicated GPU capacity
  • ✓ Fine-tunes & private models
  • ✓ SSO, audit logs, SLA
Contact sales

[ 05 ] FROM THE PEOPLE WHO SHIP ON IT

"We replaced three vendors with one Lumina endpoint. Same SDK for image, copy and voice, and our generate route went from 400 lines to about 40."

Sora Bekker

Staff Engineer, Drift Studio

"Seeds and brand-lock are the part nobody else nails. We can reproduce an asset from six weeks ago byte-for-byte. That's what made it production-safe for us."

Theo Marchetti

Founder, Pixelforge

"Streaming partials made the wait disappear. Users watch the caption type itself while the images resolve, it feels alive, and churn on that screen dropped."

Lena Aydın

Head of Product, Nimbus

[ 06 ] 200 FREE GENERATIONS · NO CARD

Start with a
single prompt.

Drop in the SDK, paste your first prompt, ship the result before your coffee's cold. The docs are short on purpose.

Start generating or read the docs →