The AI image generator in Secret Chat AI creates pictures from a description and edits images you already have. You can compare several models, change the shape and resolution, make variants, remove a background or upscale a finished picture. The price appears in dollars before you generate.
Secret Chat AI is an independent product. It is not affiliated with or endorsed by the companies behind the models named on this page. See our Disclaimers for details.
AI Image Models in Two Entries
The model picker has two image entries. Image uses the Standard moderation level. It includes GPT Image, Nano Banana and Grok Imagine. The uncensored Image generator, listed as Image Lax with a flame, uses a lighter screen. It includes Seedream, Seedream Pro, Qwen Image, Lustify and WAI Illustrious. The Upscaler is the last model in both entries.
Each family has a different character. GPT Image follows detailed instructions well and is useful when words must appear inside a picture. Nano Banana is a straightforward everyday choice. Grok Imagine can make several bold variations quickly. Seedream can draw or edit, while Qwen Image and Lustify offer more style control. WAI Illustrious draws in an anime style. When you attach a picture to WAI Illustrious, it redraws the scene from a description, so faces and small details may shift.
Try the same description on two models when the first result misses the look. A different model often helps more than adding another line of adjectives.
Settings That Follow the Model
Settings appear as chips under the message box. You may see model, preset, picture count, aspect ratio, resolution, transparent background and style. The row ends with the price of the request. Switch models and the chips change with it, so you only see controls that work.
- Aspect ratio covers square, portrait, landscape, widescreen, tall and 4:5 compositions.
- Resolution reaches 4K only on Nano Banana models. GPT Image, Grok Imagine, Seedream Pro and Qwen Image offer up to 2K. Other models use a fixed output size.
- Picture count reaches 10 on GPT Image and Grok Imagine, and 4 on Seedream, Qwen Image, Lustify and WAI Illustrious.
- Transparent background is available only on GPT Image. It returns a PNG with real transparency.
- Style presets appear on Qwen Image, Lustify and WAI Illustrious when you create a new picture. Seedream has no style preset or negative-prompt field.
The last prompt and its attached pictures stay in the box after a generation. Results appear in a grid, and the desktop slider lets you choose how many pictures sit in each row. Download the versions you want to keep.
AI Image Editing and Background Removal
Every generated picture has Download and Edit controls. Press the pencil, then give a direct instruction such as “make the coat blue” or “remove the words in the corner.” Several models accept more than one reference picture in the same edit. Each result is newly generated, so check faces, lettering and any detail that must stay exact.
Remove background uses GPT Image Flare to redraw the subject on transparency. You can choose the preset or use the eraser beneath a Standard image. The result is a PNG, though hair, glass and fine edges can change. For a full walkthrough, see the AI image editor.
AI Image Upscaler for Up to 10 Pictures
The Upscaler is a model choice, not a button or preset. Pick it in either Image entry and attach up to 10 pictures. It starts on its own, reads no prompt and opens the result in a new chat. There is no 2× or 4× choice.
A picture under 2,048 pixels on its long side is restored and redrawn at about 2K. This route can repair blur, noise and compression damage, but an AI redraw may alter small text or a face. A picture that is already 2,048 pixels or larger is enlarged two times without being redrawn. If one picture fails, the others can still arrive, and the failed one costs nothing.
Presets for Avatars, Posters and Covers
Avatar chooses a square shape and adds a short head-and-shoulders brief. Poster chooses 2:3 and leaves room for a title. YouTube cover chooses 16:9 for a channel banner and keeps the essentials in the middle band that phones show. YouTube thumbnail chooses 16:9 for a video thumbnail built to be clicked: one big expressive subject, a few huge words when you give a title, and strong contrast that reads at phone size; it arrives at 1,280×720. Facebook cover, LinkedIn cover and X header choose the widest drawing shape, 21:9; all four covers arrive in each platform's exact size, with the subject in the middle and the lower-left corner plain for the profile picture. Avatar and the covers lock their shape; Poster only starts on 2:3, and you can change it. Your own description still controls the subject and style. The AI avatar generator guide shows the square workflow in detail.
Make It Move
A finished still can become the first frame of a clip. Open the AI video generator, choose Video or Video Lax, and place the picture in the First frame slot. The video settings show the available length, resolution, shape and sound controls for the model you pick.
Limits of Image Lax
Image Lax gives lawful fiction, satire, dark themes, medical work and other creative prompts more room. Every request is still screened. Sexual content involving minors, terrorism and real threats of violence are blocked on every model. Each provider also keeps its own rules, so a request can still be refused. It means fewer needless refusals, with firm limits. The full guide to Image Lax is on the uncensored AI image generator page. For the wider picture, read the uncensored AI guide.
Pay As You Go
Image generation and upscaling use a prepaid dollar balance. There is no subscription, packs start at $12 and the balance never expires. The app shows the price of each picture or upscale before it starts, then prints the exact charge under the result. A failed generation costs nothing. The payment page explains the available packs and Volet checkout.
Privacy, Stated Plainly
Secret Chat AI stores no chat archive or generated-picture archive on its side. History stays in your browser. We build no profile of you, and model providers see our service rather than the person asking. Your material is never used for training.
The limit matters. Every description and attached picture goes verbatim to a third-party model provider, and the OpenAI classifier also sees it during screening. Provider retention may still apply. Follow-up requests on GPT Image, Grok Imagine and the Venice-hosted models are first rewritten into one prompt by a helper model at OpenAI, and a WAI Illustrious edit first sends the picture to GLM Abliterated for a description; the Session Privacy Report names every provider involved. Remove names and any other identifying details before you send them. Secret Chat AI removes you from your queries — it does not remove the data from your messages.