Image to image. A whole new style.

Choose a photo you own or have permission to edit, then select a style such as Anime, Cartoon, Manga, Chibi, Kawaii or Watercolor. The style button adds a local prompt preset and model choice to the workspace; it does not upload an image or run a model. Check the details you want to preserve before preparing the request. Public AI generation is not open in this validation build; a prepared request is not a generated image.

Yes, you can prepare multiple references for models that support them. The limit is shown beside the upload control and changes with the selected model. Images are checked locally before preview. A prompt describing each reference’s role helps separate the subject, style and composition.

Choose by the task and the controls you need. Models differ in reference count, aspect ratios, resolution, output format and optional editing controls. The model directory describes those supported settings. These are verified interface capabilities, not a quality ranking or a claim that every model has been generation-tested.

Not yet. The plans and model credit usage are a preview, and a one-time welcome trial of 10 Economy edits is planned, not issued in this build. Checkout and public generation are not open; signing in does not grant credits. Paid generation will require a verified purchase and sufficient server-authorized balance. The separate local photo editor can apply and export on-device edits now; its results need a second check before preview or download.

Use a static PNG or JPEG within the file and dimension limits shown in the workspace. Invalid, animated or unsupported files are rejected. A local content check must allow each reference before it appears. An error or a request for review cannot be bypassed.

The checked comparisons pair a licensed source photo or a project-authored digital sketch with its corresponding GPT Image2 output tested through this project’s provider adapter. Each pair identifies its source type, credit and source link. These are owner-tested provider integration examples, not customer outputs or proof that public paid generation is open. The additional input/output studies are AI-generated illustrative concepts that clearly label synthetic references and their corresponding edits. The portrait board shows four reviewed effects of the same original AI-generated reference; these are not real-photo conversion or identity-preservation proof. Unavailable or incomplete pairs are not shown as finished results.

Not necessarily. Stylization changes linework, shading, texture and sometimes proportions or small scene details. A clear prompt can describe the features to keep, but it is not a guarantee of exact identity, anatomy or pixel-level preservation. Compare the source and result carefully before using a generated image.

The current AI draft and checked references stay in page memory, not in your account history. Leaving or reloading clears them. Request preparation does not upload image bytes. A future live generation workflow must explain provider processing and retention before you submit.

A failed or uncertain content check holds the image instead of releasing it. The current preparation flow explains validation errors and unavailable execution. It does not start a paid task, invent a result or automatically retry a provider. Live failure, settlement and recovery terms must be defined before purchasing is opened.