Image and video decision
PhoneDiffusion
A technical user can build a narrow local iOS image generator using existing open-source engines and model toolchains, but reproducing the full polished App Store product (model catalog, compatibility detection, UX polish, in-app commerce) is substantial and time-consuming.
View on the App Store↗$9.99/mo
$120/yr
Read off the official pricing page.
$100one-off160 h to build
$0/mo4 h/mo upkeep
On cash alone, building overtakes the subscription at 1 seat.
No open-source build does this yet
Nothing published replaces this one, so a replacement starts from an empty file. Here is what it would have to cover.
What a replacement has to do
- Load a compatible diffusion model to device, enter or dictate a text prompt, run on-device generation, present live progressive output, allow inpainting/img2img/upscale edits, and save results to a private gallery on the device.
What it still won’t have
- Polished App Store distribution, in-app purchase plumbing and review handling
- Device-specific optimization and thermal/memory tuning performed by the vendor
- Built-in curated model collection and compatibility detection across many iPhone/iPad models
- Polished UX features such as live generation tuning, localization and ongoing UI refinements
What remains hard
- Execution quality
Fast as hell, tiny models, pretty much the best local so image generator out there on iOS.
First-year cost
Keep paying
Paying is—cheaper in year one.
On cash alone, building overtakes the subscription at 1 seat.
Money you would actually spend
Time you would spend
—
What you would spend
What we assumed
The verdict above measures whether you could build it. This one is only about money.
Runnable build prompt
Build a minimal local iOS image-generation app in Swift/SwiftUI that runs Stable Diffusion models on-device using Core ML / Metal or a local runtime (e.g., integrate LocalAI or a converted Core ML model): include model download/verification, a prompt input with optional voice dictation, live progressive generation preview with pause/accept, inpainting brush and img2img upload, basic 4x upscaling, a private on-device gallery storing prompts/settings, and share/export. Out of scope: App Store in-app purchase plumbing, multi-model marketplace, server-side inference. Include error handling, device compatibility checks, background model download, and unit/UI tests for core flows.
How we checked
How the score was reached
- Partly verdict base52
- Price verified on pricing page+3
- 3/3 assessment runs agreed+4
- Evidence score59
The base comes from the verdict. Everything under it is a check that either happened or did not, and each one is a fact frozen in this record rather than a judgement made at render time - so the same evidence always produces the same number.
How scoring works →Integrity checks
What held up, and what did not.
