Create polished stills, readable short typography, reference-guided edits, product mockups, and high-quality campaign drafts.
Reference gallery examples from Zanta illustrate visual directions. They are not results verified on this site.
Bring short words into the composition with an explicit reading order, consistent spacing, and room for the main image.
Turn a written explanation into a visual board, cutaway, or short illustrated sequence that can be reviewed at full size.
Show a product in a coherent visual world with deliberate materials, light, and clear copy placement.
Explore a dashboard layout or a sequence of panels with a repeatable palette and recognizable subjects.
Use up to four visual references to describe the subject, setting, or finish you want to carry into a new image.
New starting briefs for the visual directions shown below. The images are reference gallery examples from Zanta, not verified outputs of these prompts.
Design a square illustrated night-market poster. A lively but uncluttered central street scene with warm lanterns, cream paper, midnight blue shadows, and orange accents. Exact text only: NIGHT MARKET, FRIDAY 7 PM, RIVER WALK. Give the title strong hierarchy and put the time in a separate readable footer. No logos or extra words.
Create a delicate paper-cut illustration of one amber butterfly above two sage-green leaves. Fine layered edges, soft studio light, a clean isolated silhouette, no words or logo. Ask for a transparent background where available, and keep the subject separated from any background for later cutout work.
Design a clean desktop storefront concept for a fictional ceramic studio. One large cream mug hero and three small product cards, warm ivory background, restrained brown typography, generous spacing. Exact visible words only: CLAY STUDIO, MUG, BOWL, VASE. Keep it a visual interface concept rather than a functioning website.
Create a four-panel wordless visual story of one adult wearing a yellow raincoat and carrying a red umbrella beside a city tram. Panel 1: waiting in rain. Panel 2: tram arrives. Panel 3: stepping aboard. Panel 4: looking through the window. Keep the face, coat, umbrella, and palette consistent in all panels.
Write a production brief that makes the result easy to inspect, compare, and refine.
Start with the poster, product still, character sheet, diagram, or campaign image you actually need.
Explain where the subject, copy space, supporting objects, and focal point belong in the frame.
Name surfaces, light direction, lens feel, texture, environment, and the intended finish.
Quote the required words exactly and review spelling, casing, punctuation, and invented copy.
Call out fake logos, extra subjects, source drift, broken anatomy, or an unsafe crop before generating.
Choose the model by the job and review the visible failure mode before you publish.
Compare the controls available in Dola AI before opening another workspace. Each model keeps its own identity, settings, and credit estimate.
| Model | Inputs | Output settings | Workspace |
|---|---|---|---|
| GPT Image 2.5 | Prompt + up to 4 images | 1K / 2K / 4K | Open model |
| Nano Banana | Prompt + up to 4 images | 1K | Open model |
| GPT Image 2 | Prompt + up to 4 images | 1K / 2K / 4K | Open model |
| Seedream 5 | Prompt + up to 4 images | 2K / 4K | Open model |
| Grok Image 2.0 | Written prompt | 1K / 2K | Open model |
Keep the first pass comparable, then spend more only after the direction is working.
Define the asset, audience, format, and one visible outcome.
Describe the subject, composition, material, light, motion, and reject conditions.
Select only the ratios, sizes, quality tiers, and references supported by this route.
Inspect the complete result, then change one variable for the next attempt.
Public walkthrough links. Open a video on YouTube to review its context, author, and date.
A public video linked from the reference gallery. Opens YouTube in a new tab.
External linkA public video linked from the reference gallery. Opens YouTube in a new tab.
External linkA public video linked from the reference gallery. Opens YouTube in a new tab.
External linkExplore independent community discussion. These links open external sites.
Follow the provider directly for current announcements. Availability here is shown in the workspace above.
Practical answers about access, settings, references, review, and output use.
Sign in, write a concrete visual brief, choose the available output settings, and check the credit estimate before generating. Your completed images appear in Assets.
Eligible accounts can use the credits currently available in their balance. Daily check-in may add credits when enabled. Each generation uses the amount shown in the generator; access is not an unlimited free offer.
Add up to 4 JPG, PNG, or WebP images. Describe what each image should guide and what needs to remain recognizable. Review the output rather than assuming exact preservation.
The generator offers 1K, 2K, 4K output and 1:1, 4:3, 3:4, 16:9, 9:16, 2:3, 3:2, 21:9, 5:4, 4:5 framing. Choose from the controls shown and check the updated credit estimate.
Quote the exact words, keep them short, describe their placement and hierarchy, and review spelling and punctuation at full size. A visually convincing layout may still contain text mistakes.
Long or conflicting instructions make a result harder to control. Keep one main subject, state the protected details, and change one visible variable at a time when refining the image.
These are examples from the reference website, reused for visual comparison. The prompt examples are newly written starting briefs, not the original prompts or a claim that this site reproduced the displayed images.
Check text, anatomy, product geometry, factual labels, and the final crop. Review the current site terms and make sure you have permission to use any supplied references or recognizable material.
Use GPT Image 2.5 for product concepts, short typography, structured illustrations, visual stories, and reference-guided edits that benefit from an explicit composition.
This page offers the configured GPT Image 2.5 model and its available output sizes. It does not present Flare or Sunburst as separate products.
There is no dedicated transparency switch in this generator. Describe an isolated subject and inspect the returned background; use the background-removal tool when you need a separate cutout.
Use the same subject, format, and visible constraints across models. Compare text accuracy, structure, source fidelity, and the credit estimate rather than assuming one model suits every job.
Create polished stills, readable short typography, reference-guided edits, product mockups, and high-quality campaign drafts.
Try the generator