How to Create Images With AI, Arabic Text Included: From Prompt to Video

01How to create images with AI when the text is Arabic

If you have ever asked an image generator for a Ramadan sale poster, you have probably seen the result: disconnected letters, invented words, or Arabic running left to right. Many models treat Arabic as shapes to imitate rather than letters to write. Vocla's image studio was built around one rule that fixes this: anything you write between quotes is passed to the model letter for letter and drawn exactly as typed.

The studio lives on the Images page in the web app's sidebar. It has four modes, picked from tabs at the top of the composer. This guide walks through them in order, from your first prompt to a finished video.

02Write a prompt the model understands: five parts

A good prompt reads like a clear brief to a designer. No jargon needed; write the way you would explain it to a person. Cover these five parts:

1. 1Subject: what exactly is in the picture? A brass dallah, a box of premium dates, a man in a red shemagh.
2. 2Setting: a warm majlis, white marble, dunes at sunset, a bright studio.
3. 3Style: realistic photo, flat illustration, 3D render, gold Arabic calligraphy.
4. 4Light and colour: soft golden light, or your brand palette such as navy and gold.
5. 5Text in quotes: the headline, price or name, plus where it sits in the frame.

Then choose a preset: YouTube thumbnail, Instagram post, Story and Reels, podcast cover, product photo, poster and ad, logo and sticker, or freestyle. The preset sets a sensible shape, which you can override: square 1:1, portrait 4:5, story 9:16, wide 16:9 and more. You can make 1 to 4 images per run to compare options side by side.

03Quotes: the trick for exact Arabic text

This is the single most useful thing in the guide. Text «in quotes» is not interpreted or reworded; it is drawn as typed, with joined letters, hamzas and numbers intact. The composer has an Add text button that wraps your words in quotes for you. Because the text is literal, any typo in your prompt will appear in the image, so proofread first. If you don't read Arabic, have someone who does check the quoted words.

• Keep it short: a 2 to 5 word headline renders far more cleanly than a paragraph.
• Say where: at the top, on the shop sign, centred in large type.
• Describe the lettering: gold calligraphy, bold modern sans, thuluth, to match the mood.
• One quote per line of text: put the headline and the price in separate quotes and place each.
• Numbers as you want them: write the price exactly as it should appear, such as «99 ريال».

Word choice matters too. Social ads in the Gulf often speak the audience's dialect, while print and formal material stays in Modern Standard Arabic. The same offer, two registers:

04Which quality: Lite, Vox Image, Pro or Ultra?

You pick a quality before generating. All four write Arabic; the higher ones keep long text sharper. Credits below are per image at 1K on current settings:

A practical rule: test the idea and composition with Lite, move to Vox Image once the direction is right, and switch to Pro when there is a long headline or you need a logo with a clear background. Save Ultra for large prints and campaigns that matter. Images that fail cost nothing; the credits come back.

How many images will you get? Divide your plan's credits by the per-image cost. Creator's 15,000 credits make about 37 Vox Image pictures at 1K, or about 125 with Lite. Starter's 5,000 credits make about 12 Vox Image pictures. Remember that the same credits also pay for voice and video.

Vox Image runs on Google's Nano Banana 2.1, part of the Gemini Image model family per Google DeepMind's model page as of October 2026 (deepmind.google/models/gemini-image). Vox Image Pro runs on an OpenAI GPT Image model. Credit rates are current defaults and can change; the studio shows the exact figure before you generate.

05Edit a photo with one sentence

Edit is the mode most people end up using, because they already have a photo and want to change one thing. Upload a photo or pick one of yours, describe the change in a sentence and press Apply edit. The viewer shows before and after so you can drag to compare.

The second half of each line does the heavy lifting: say what must not change. «Keep the face as it is» or «keep the label» gives a far more faithful result. You can also add reference photos to bring things in and merge them: up to 4 on Creator and 10 on Professional and above, for example «put the person from photo 1 next to the car in photo 2».

06Remove the background and enhance quality

Two one-click tools replace whole apps. Remove background takes one photo and returns the subject alone on a transparent PNG, in the same shape as your photo and never cropped, for 300 credits on current settings. Use it for product listings, stickers and placing a person or product on a new design.

Enhance returns the same photo sharper and cleaner, removing blur, noise and compression artefacts without adding or changing anything: 2K for 400 credits, or 4K for 700 credits on plans with 4K. It is ideal for old photos, soft phone shots and anything headed to print.

07From image to video, and ideas by use case

Any image in the studio can become a video. Press Make it a video to send it to Vocla Video as the hero photo of an ad with an Arabic dialect voiceover and correctly shaped Arabic captions, or drop it into your own clip in Video edit. On current plans, the ad from a photo starts on Starter; plans without it show an upgrade prompt.

The most common uses, each with its own detailed guide on the blog:

• Seasons: Ramadan and Eid greeting cards with your name or brand, in story or post sizes.
• Online stores: product shots on white or in lifestyle scenes, then a video ad.
• Creators: YouTube thumbnails with a clear Arabic headline, and podcast covers.
• Brands: logo ideas with an Arabic name on a transparent background, and ad designs.
• Events: wedding and party invitations with names spelled right.

Prices show in your local currency and are charged in USD. Per-plan settings can change; the studio always shows the current figure before you generate.