macOS · bring your own keys · pay once

Fire a prompt.
Compare what comes out.

Kiln runs one prompt across every image model — or many phrasings against one — then lays the results side by side, so you keep only what survives the firing.

Pre-order $9.99 · once ▶  Watch the demo
Native macOS app No subscription Refundable until launch
Kilnuntitled ⚙
Run · 028
“A copper-wire sculpture of a hummingbird mid-flight, studio lighting, deep navy backdrop, shallow depth of field”
GPT Image 2 Nano Banana Pro Flux 1.1 [pro] Nano Banana
Outputs4 models · 1:1
A.GPT Image 2t 6.2s$0.053
Example render from GPT Image 2output.png1024×1024
B.Nano Banana Prot 4.6s$0.134
Example render from Nano Banana Prooutput.png1024×1024
C.Flux 1.1 [pro]t 5.1s$0.040
Example render from Flux 1.1 prooutput.png1024×1024
D.Nano Bananat 3.8s$0.039
Example render from Nano Bananaoutput.png1024×1024
3 providers 6 models in one run OpenAI · Google · Fal keys live in macOS Keychain direct calls — no proxy, no telemetry
Watch · 38 seconds

See a firing, start to finish.

One prompt, four models, side by side — in the time it takes to read this sentence twice.

The whole point

Comparison is the feature.

Most tools generate one image at a time. Kiln holds everything constant and changes exactly one thing — so a comparison actually means something. Two ways to do it.

01

Compare models

one prompt → many models

Send a single prompt to every model at once. Same words, same aspect ratio, every provider in parallel. The only thing that changes is the model — so the grid shows you exactly how each one reads you.

1 promptup to 6 models1 click
02

Compare prompts

many prompts → one model

Pin one model and pit phrasings against each other. Four ways to describe the same idea, one model held constant, a fair test every time. Find the wording that actually lands.

up to 4 prompts1 modelheld constant
Built like a tool, not a toy

Quiet on the surface.
Serious underneath.

Bring your own keys

Your OpenAI, Google and Fal keys, stored in the macOS Keychain. Kiln calls each provider directly — there's no middle-man server.

keychain · direct

Local-first & private

Runs and outputs live on your disk. No account, no cloud, no telemetry — your prompts and images never leave your Mac except to reach the model.

your machine

Reference images

Drag an image in, or paste with ⌘V. Kiln routes it to every model that accepts references, so img2img comparisons run side by side too.

drag · ⌘V

Cost, before you commit

Every model shows its estimated price per image before you run, and each card reports what it actually charged after. Know what a firing costs.

$ / image

Lightbox compare

Click any result to fill the screen, then step through the grid full-size. Spot the difference your thumbnails were hiding.

fullscreen

Every run is kept

Kiln files each firing automatically — prompt, settings, outputs and cost. Scroll back through your history and re-open any run exactly as it was.

local history
Pricing

Buy it once.
It's yours.

No seats, no subscription, no usage meter skimming your runs. You pay the providers directly for generation; you pay Kiln exactly once for the workbench.

Founder price, locked forever All providers
founder Kiln for macOS
$9.99 one‑time
$39 at launch · pre-order saves you the difference
  • Every provider & model — OpenAI, Google, Fal
  • Unlimited runs & a local history
  • Both compare modes — models & prompts
Pre-order Kiln $9.99
one-time · delivered when 1.0 ships · refundable until launch
Kiln
Bring your own keys · keep your own work

Stop guessing.
Start comparing.

Fire the same idea at every model and let the grid settle the argument. Pre-order Kiln at the founder price and run your first comparison the day it ships.

Pre-order $9.99 · once Watch the demo
macOS 14+ Apple silicon ships at launch