A new app from Latentry
Generate images.
Entirely on your iPhone.
Latent Canvas turns a text prompt into a high-quality image using a state-of-the-art model that runs entirely on your device's GPU — no cloud render farm, no image uploads, and it works in airplane mode.
Real image generation, running on your phone
Most AI image apps send your prompt to a server and stream a picture back. Latent Canvas runs the whole model locally — the image is drawn on your iPhone.
On-device generation
A real diffusion model runs on your iPhone's GPU — no cloud render, no waiting in a queue.
Works offline
Create in airplane mode, on a plane or off-grid. No connection needed to generate.
No image uploads
Your pictures are made and kept on your device. Finished images are never sent to a server.
Unlimited, no credits
The free tier generates as much as you like — no per-image credits, tokens, or daily caps.
Higher resolution
Premium unlocks larger, higher-resolution output for crisper, more detailed images.
Save to Photos
Keep what you make — save creations straight to your photo library and share anywhere.
Your art is made on your device
There's no server drawing your pictures — the model runs on your iPhone, and the images you create are never uploaded. Because Latent Canvas is an open image generator, keeping it safe means you sign in with Apple and your prompts are screened for prohibited content (on your device and on our servers) before an image is made. Everything else stays on your phone.
See exactly how sign-in, prompts, and purchases are handled in the Latent Canvas privacy policy →
Free to start. Premium for more.
Generate unlimited images free at the standard resolution. Premium unlocks higher resolutions — choose monthly or save with yearly.
Free
Unlimited generation at standard resolution. No credits, no daily caps, no account paywall on the core experience.
Premium — $4.99/mo or $49.99/yr
Unlock higher-resolution output. Yearly works out to two months free. Cancel anytime in Settings.
Requires iPhone 15 Pro or newer (8 GB). On-device generation uses the device's GPU and a one-time model download.