A Google AI image generator can mean image creation inside Gemini or inside Google Search. Both have documented consumer workflows, but their entry points, account requirements and limits should be checked separately. For a new illustration you intend to refine through conversation, Gemini is a practical place to begin.
Choose the surface before choosing a model name
| Your task | Surface to investigate | Main boundary |
|---|---|---|
| Create a picture and revise it through conversation | Gemini Apps | The account, available image model and current quota determine access |
| Create an image while using Google Search | Search / AI Mode | Search has its own eligibility and daily-limit guidance |
| Integrate generation into your own software | A developer API | Consumer instructions do not establish API billing or implementation |
| Build an original interactive companion | Cherrypop’s character creator | Appearance is one part of a character that also has personality and a scenario |
Nano Banana and other model names do not answer the whole access question. A tutorial can describe a model correctly while sending you to an interface with different controls. Record the product surface first, then the model shown inside it. That habit also makes later troubleshooting more precise.
Start an image in Gemini
Google’s Gemini image help describes signed-in creation through the image tool and editing either generated or uploaded images. Its current guide names Nano Banana 2, with a separate paid-subscriber Redo with Pro option. Country, language and account restrictions apply. Use the current image tool offered to your account rather than copying a model selector from an older screenshot.
For example, plan a fictional reading corner with a green armchair, a floor lamp and a clear walkway. State that the image is a concept illustration. Do not describe it as a measured room design or ask the image to stand in for dimensions you have not supplied.
| Requirement | Give the generator | Check in the result |
|---|---|---|
| A readable layout | One chair beside one lamp, with an open walkway | The furniture does not obstruct the proposed path |
| A coherent treatment | Warm evening light and a restrained color palette | The scene fits the intended mood |
| Room for a caption | An uncluttered lower-left region | The caption can remain legible at the final crop |
| A bounded revision | Change the chair upholstery while retaining the layout | The lamp, window and viewpoint do not change unintentionally |
Use the worksheet to make acceptance visible. If you cannot point to the difference between a usable version and a rejected one, another generation is unlikely to settle the decision.
After creating a baseline, request one meaningful edit. Google documents reopening an image from Library and continuing through Chat; Download full size is the documented saving control. Save the accepted version and inspect the file itself. Account-specific download options are more useful than assuming a promotional preview establishes your export entitlement. Gemini editing and download instructions
Use Search image creation when that is your starting point
Google’s Search image guide describes creating an image from a search prompt or selecting Create Images in AI Mode. It requires an eligible personal account and supported location/language. Image edits are requested through the conversation. The guide publishes daily limits by plan and warns that capacity and demand can change them.
If you already have a visual idea while researching a subject, this route can keep the creation task near that context. If you want to iterate across a set of visual assets, first decide where you will retain the approved versions. Switching surfaces mid-project can make it harder to know which quota, model or conversation produced an image.
Diagnose the missing control, not the whole product
| What happens | Check first | Avoid concluding |
|---|---|---|
| No image option appears | Account eligibility, language and product surface | That every Google image tool is unavailable |
| Generation stops after earlier attempts | The current limit notice and reset guidance | That the same limit applies in a separate product |
| An edit changes too much | Whether the request states both the change and preserved details | That the requested object was guaranteed to stay exact |
| The file is unsuitable for delivery | Actual dimensions, crop, wording and terms | That the attractive preview is a finished asset |
Treat tool access and output suitability as separate checks. A paid account can unlock a feature without making every image accurate. Likewise, an unavailable control does not prove a hidden workaround is appropriate; choose a documented route that your account supports.
Google’s image-generation overview describes visible and SynthID markings. Do not treat a generated image as a documentary photograph simply because it looks convincing. When accuracy matters, verify its factual content independently and identify the use of generated imagery appropriately.
When the image is the start of a character
A general image tool helps make a visual. If you want the subject to become a companion you can interact with, Cherrypop’s creator offers a character-first route: choose appearance and style, then define personality, scenario and an opening message. That context gives the character something to do beyond posing in an image.
Cherrypop also provides image and video creation entries, but we do not claim it duplicates Google’s editing controls or preserves every visual detail across outputs. Choose it for the companion workflow; choose a general image editor when your main deliverable is a room concept or a layout asset.
It is free to start with limits, and some media features require Premium or Cherries. If an original character is your goal, create its appearance and personality and check the displayed generation cost before making media.