In short
Start with one clear scene, describe its subject and composition, then improve one detail at a time. Check Persian lettering separately and keep sensitive photos out of practice uploads. Before choosing a tool, verify access and usage limits: support for Farsi does not establish availability inside Iran. [1][5][7]
Key takeaways
On this page 13 sections
- How do you turn an idea into a useful AI image?
- What should you check before choosing a tool?
- How can you write a clear Persian prompt?
- How do you improve a result without starting over?
- Why does Persian text need a separate proofread?
- How can you keep the scene consistent?
- Which photos and documents should you avoid uploading?
- Can you use the picture for a shop or paid project?
- What should you save before publishing?
- Common mistakes
- Frequently asked questions
- Glossary
- Sources
How do you turn an idea into a useful AI image?
Describe one clear scene, choose its shape and visual style, then revise the result with a specific follow-up request. A useful prompt tells the tool what should appear and how it should be arranged. Official image guides recommend concrete subjects, composition and lighting rather than a pile of impressive-sounding adjectives. [5][4]
For example, begin with a picture of a handmade ceramic bowl for a shop announcement. Decide whether you want a product photograph, an illustration or a decorative background. Keep those purposes separate: a pretty invented bowl is not a faithful photograph of the item a customer will receive.
This guide is current as of November 4, 2026. The sample prompts below are original writing examples, not tested outputs or promises about a particular model. They are in English here so you can see the structure; use the same structure when writing your own Persian instructions.
Before spending money, check whether the service supports your location, account and intended use. In Gemini, Farsi appears on the supported-language list, but Iran does not appear on the web app's supported-country list. Language support and access from inside Iran are different questions. [1]
What should you check before choosing a tool?
Start with access, privacy and the task you actually need. Do you want a new illustration, a change to your own photograph, or lettering for a finished poster? These are different jobs. Gemini's image help covers generation and editing, but eligibility and available features depend on the account and service conditions. The workflow here is intended for adults. [2]
Open the provider's own product and help pages before trusting a reseller's offer. Check the country list, sign-in requirements and the controls visible in your account. A screenshot from someone abroad does not show what an account in Iran can use. Do not pay a stranger merely because they promise an unavailable feature will work.
Budget for experiments. Usage limits can change, and Google's current limits page describes limits that depend on the model and work being requested. It does not justify promising every reader a fixed number of free images each day. Check the limit notice in your own account before starting a deadline-sensitive job. [10]
Write a small acceptance checklist first: correct object, suitable crop, readable text and permission to use the material. That gives you a way to decide when the picture is finished instead of generating variations without a clear goal.
How can you write a clear Persian prompt?
Build the description around a subject, an action or state, a setting, a composition and a style. Google's prompting guidance uses these kinds of details to make the intended picture clearer. You do not need to imitate technical camera language you do not understand. Plain descriptions such as close view, soft window light and empty space above the object communicate a practical brief. [5]
Try this original example: “A single turquoise ceramic bowl on a simple wooden kitchen table, soft morning light from the left, viewed slightly from above, realistic product photograph, square composition, open space around the bowl.” Translate the description naturally into Persian if that is how you prefer to work. Keep the object and arrangement consistent when comparing attempts.
For an illustrated event background, change the purpose deliberately: “A flat illustration of a pomegranate branch along the lower edge, warm cream background, two muted colors, wide empty area in the center for a title.” Add the title separately if exact lettering matters more than generating everything together.
Adobe also recommends direct, descriptive wording and revising the prompt when the result misses the idea. Choose details that affect the picture, rather than adding a long list of unrelated styles. [4]
How do you improve a result without starting over?
Make the next request describe the change you want. If the object is right but the background is wrong, ask for a different background while keeping the object, angle and framing. Image-editing workflows support follow-up changes, although a request to preserve a detail is not a guarantee that every pixel will remain identical. [2]
An original follow-up might be: “Keep the bowl and its position. Replace the patterned cloth with a plain cream surface.” Compare the new image with the previous version. Check the rim, handles, pattern and proportions, not just the background you asked to change.
When using several reference images, give each a clear role. One might show the object, another the intended arrangement and another the color mood. Google's guide recommends explaining what each reference contributes. Do not assume the tool will infer which features are essential to your product. [3]
For your own working process, keep a short note of what changed between versions. If one attempt fixes the lighting but damages the object, return to the stronger version and make a narrower request. This is a way to organize your decisions, not a claim that repeated prompting will always solve the defect.
Why does Persian text need a separate proofread?
An image that looks polished can still contain incorrect lettering. Google's image guidance lists text clarity and accuracy among limitations that can remain. Treat generated words as a draft to inspect, especially names, prices, dates and contact details. A believable-looking sign is not evidence that its spelling is correct. [3]
Persian text also has layout requirements. The W3C's Arabic and Persian layout draft describes right-to-left writing with numbers running left to right. That makes mixed text, dates and Latin brand names worth checking carefully. The document explains script layout; it is not a certification of any image model's Persian output. [6]
Copy the intended wording into a separate note before generating. Compare the result character by character, then read the full line naturally. Ask another Persian reader to check a public-facing announcement when possible. For a price, compare the digits and currency wording against the approved product information.
If lettering keeps failing, use the generated image as a background and add the text in a layout editor with a font you can inspect. Leave space for that text in the original composition. This also lets you correct a date later without regenerating the entire picture. Do not publish an unreadable result simply because its decoration is attractive.
How can you keep the scene consistent?
Describe the visible outcome you want and keep the brief focused. Google's Cloud prompting guide recommends a clear subject, context and composition, with changes refined through iteration. Those principles can help organize a prompt, but model-specific controls in a developer product should not be assumed to exist in every consumer app. [5]
For a small shop, write down which details must stay true: the number of pieces, their color, the position of a label and whether accessories are included. Separate those requirements from optional choices such as wall color or lighting mood. Compare the result against that list before showing it to a customer.
For a fictional illustration, consistency has a different purpose. You might want the same jacket color across three scenes or the same room arrangement in a sequence. Save a chosen reference and describe what should carry over. Reject a version that changes a defining feature, even if the new picture is more dramatic.
Keep the final review tied to the intended use. An imagined room can be suitable for a mood board; an advertisement should not imply that invented furniture comes with the real product. When a picture explains a process, have someone who knows that process check it. Visual plausibility alone cannot establish factual accuracy. [3]
Which photos and documents should you avoid uploading?
Use a non-sensitive sample while learning. Google's privacy notice says some collected data is reviewed by people and warns against entering confidential information you would not want reviewed or used to improve services. Turning off an activity setting does not mean the service stops processing everything you send. [7]
For readers in Iran, make a deliberate decision before uploading an identifiable face, private conversation, identity document, workplace record or photograph that reveals someone's location. Ask whether the task needs the real material at all. A fictional product brief or a photograph of an ordinary object may be enough to learn the workflow.
Inspect the whole frame before uploading a reference. Look for a visible address, a badge, a computer screen, a reflection or another person in the background. Removing unnecessary information before upload is easier than trying to undo disclosure afterward. Keep the original private file separate from the copy prepared for the task.
Get the person's permission before using their face in an edit, and agree on the intended audience and purpose. Google's use policy prohibits privacy violations and deceptive impersonation. A tool accepting an upload does not prove that the person pictured agreed to the use. [9]
Can you use the picture for a shop or paid project?
Check the rights to the input material and the service's current terms separately. Copyright permission may be needed for photographs, illustrations or other protected work you upload or reproduce. WIPO explains that rights and exceptions depend on the applicable law; finding an image online does not automatically give you permission to reuse it. [8]
For a practical rights record, keep the original files you own, any license you relied on, the agreed use and the date. If a client supplied a photograph, ask who created it and whether the permission covers this project. Do that before promising a public campaign or a resale product.
Do not make an image that falsely presents a real person as endorsing a product or taking part in an event. Google's policy addresses deceptive impersonation and misleading provenance. For a clearly fictional illustration, choose a presentation that does not invite readers to mistake it for documentary evidence. [9]
If the project involves a recognizable person, a valuable brand or disputed ownership, get advice for the relevant jurisdiction before release. This guide cannot decide a particular license dispute. A paid subscription is a service purchase; treat any rights promise as something to read carefully rather than an automatic answer to every third-party claim.
What should you save before publishing?
Review the chosen image at the size where people will actually see it. Read every word, inspect the main object and compare it with your brief. Check that a crop does not cut off a label or change the meaning. Ask whether the image could mislead a buyer or a reader about something real.
Download the finished file through the available image controls. Google's help describes downloading generated images, but output options depend on the product and plan. Inspect the actual file you receive instead of relying on a promotional claim about maximum quality. [2]
Keep a project folder containing the prompt, the selected image, your approved wording and any permissions. Add a simple note explaining the intended use and the details you checked. If you need to revise the picture next month, this record will be more useful than a collection of unnamed screenshots.
You can also explore PYM Studio when planning a visual project. Check the tools and conditions shown there before uploading anything. This guide does not promise that a particular outside model is included, free or accessible from Iran. Finish with a picture that meets your purpose, rather than continuing until a usage limit stops the session. [10]
| Part of the brief | Example choice | Review question |
|---|---|---|
| Subject | One ceramic bowl | Is it the intended object? |
| Composition | Square, space around the object | Will the intended crop work? |
| Lighting | Soft light from the left | Does it fit the visual brief? |
| Lettering | Add approved text separately | Are every word and digit correct? |
Common mistakes
Adding many conflicting styles
Fix: Choose one visual direction before asking for variations.
Treating fluent-looking Persian as correct
Fix: Compare the lettering with the exact approved wording.
Uploading a private document as a test
Fix: Use a harmless sample that teaches the same workflow.
Assuming a paid account clears every right
Fix: Check the input license and intended use separately.
Frequently asked questions
Must I write every prompt in English?
No universal English-only rule follows from these sources. Gemini lists Farsi among supported languages. Write a clear brief in a supported language and judge the actual result. [1]
Does Farsi support mean Gemini is available in Iran?
No. As of November 4, 2026, Iran is absent from the official Gemini web app country list. Check location and account eligibility separately. [1]
Can I fix only the background?
You can request a targeted edit. Compare the result carefully because other details may also change. Keep the previous version if it is closer to your brief. [2]
Glossary
- Prompt
- The written instructions describing the image you want.
- Reference image
- An image supplied to guide a subject, arrangement or visual style.
- Aspect ratio
- The relationship between the width and height of an image.
- Usage rights
- Permissions governing how particular material may be used.
Sources
- Where you can use the Gemini web app Google
- Generate & edit images with Gemini Apps Google
- 7 tips to get the most out of Nano Banana Pro Google · 2025-11-20
- Writing effective text prompts Adobe
- The ultimate Nano Banana prompting guide Google Cloud · 2026-03-06
- Arabic & Persian Layout Requirements — Group Draft Note W3C · 2025-10-02
- Gemini Apps Privacy Hub Google
- Frequently Asked Questions: Copyright WIPO
- Generative AI Prohibited Use Policy Google
- Gemini Apps limits & feature availability Google
Comments
Top comments