shape a song with our dreampool.

LOOKING GLASS AI · SEE INTO YOUR FUTURE

Make something remarkable. Pay only when you make it.

Turn your photos into polished portraits, build original songs, and shape new worlds in the Mirage Factory. No monthly meter. Pay once for a package and keep everything you make.

Only pay for what you use30-day private galleries on every packageYour finished work is yours
Artwork of one woman seeing executive, explorer, and fantasy versions of herself in an ornate mirror
One face. Many possible futures.

From selfies to somewhere else.

Choose a pack, check that you have 15–20 varied selfies, and get a private gallery built around a personal model of your face.

01

Choose a pack

Pick the number of finished portraits and distinct scenes.

02

Check your selfies

Use the built-in checker to prepare 15–20 strong photos.

03

Train your model

Your face becomes the reference for this portrait order. Measured on our NVIDIA Tesla P40: a 192-step SDXL training benchmark took about 27.4 minutes (3 of 3 runs).

04

Choose favorites

Review a private gallery and download the portraits you love.

A lot of portraits for not a lot of money.

Every pack includes personal model training and a private gallery. The middle pack delivers 48 finished portraits across eight scenes for $37.

AI-generated example of a polished professional portrait

Starter

$27

A focused first look for headshots, profiles, and one strong transformation.

  • Finished portraits24
  • Distinct scenes4
  • Personal modelIncluded
  • Files kept30 days
AI-generated example of an elaborate celestial fantasy portrait

Premiere

$67

The full transformation: twelve worlds and enough finished work for a whole campaign.

  • Finished portraits96
  • Distinct scenes12
  • Cost per portrait70¢
  • Files kept30 days

ONE PURCHASE · NO SUBSCRIPTION

Your gallery waits for you. Not the other way around.

Buy a package when you need it. Pay once, keep what you create.

Three packages from $27
Direct purchase in USD. Pay once for the package you choose.
Starter24 portraits · 4 scenes · files kept 30 days
$27
Studio48 portraits · 8 scenes · files kept 30 days
$37
Premiere96 portraits · 12 scenes · files kept 30 days
$67

TEST checkout — nothing is charged until the merchant account is live.

THE HONEST MATH

No subscription draining money between ideas.

If you make fewer than five songs in a month, pay-as-you-go costs less than the usual subscription. If your pace ever makes a subscription the smarter deal, we will tell you plainly.

The promises

  • Every package (Starter, Studio, Premiere): your uploaded selfies, trained model, and generated gallery are kept for 30 days from the order date, then deleted. Order and payment records are kept as the law requires.
  • Failed jobs refund automatically
  • You own what you make
  • Use finished work commercially

PORTRAIT ORDER CHECK

Prepare your Studio order

Studio portrait pack48 portraits · 8 scenes · personal model
$37
Choose your portrait style

Sets the wardrobe and styling of every scene in your gallery. Mixed gives you half masculine-styled and half feminine-styled scenes.

Choose 15 different photos of yourself. Mix the angles, the lighting, and your expressions. Not 15 quick selfies taken in the same bathroom — near-identical shots train a weak model.

Every photo:

  • Exactly one face — yours
  • No sunglasses or hats
  • No heavy filters

No photos selected yet. Images are resized on this device before upload.

You caption every photo yourself. That text becomes training data for your personal LoRA, so describe only what is actually visible. Separate every detail with a comma. We do not auto-caption with AI because machine captions can be wrong.
Example: “person, looking at camera, blue shirt, tan pants, smiling”

Add 15–20 photos to begin captioning.

Every selected photo needs your caption before the order can be submitted. Checkout runs in TEST mode until the merchant account is live: you confirm a test payment, nothing is charged, and your order goes straight to the training queue. Measured on our NVIDIA Tesla P40, a 192-step SDXL training benchmark took 1644.7, 1645.0 and 1643.5 seconds (about 27.4 minutes per run, 3 of 3 runs). Training is the first stage only; a full package from upload to finished gallery has not been timed, so we do not quote a delivery time.