CREATEVISION AI · WORKFLOW TESTS
Which output is closer to the ad you need to publish?
Tested 6–7 November 2026 · three repeats per tested configuration.
For this exact-copy ad brief, start with Flare Low. All three initial posters and three revised full-brief outputs visibly include the requested headline, offer and prices.
Medium’s quote is higher: 450 credits for three full-brief requests versus Low’s 210. This small task does not establish a copy-compliance reason to upgrade.
Midjourney supplies more visual alternatives, but none of the 24 poster-stage outputs contains the complete required advertising copy in this batch.
Separate visual exploration from copy approval.
First compare the text-free visual for composition and useful space. Then check the poster word by word: NOMA CITRUS, SUMMER EDITION, $12 and 2 FOR $20. The revised brief asks for $10, 2 FOR $18 and cool evening light. A convincing bottle does not excuse the wrong offer.
- Midjourney’s text-free outputs explore several bottle shapes, camera distances and stone arrangements. The Flare outputs stay closer to an upright drink bottle with room above. We would use these as different creative directions, not award an aesthetic score.
- For the text-bearing task, Low and Medium both render NOMA CITRUS, SUMMER EDITION and the requested offer across the three repeats. Low is the sensible first trial for this brief. Both introduce bottle graphics and foliage that were not specified, so correct copy alone does not mean every detail obeys the brief.
- The Midjourney poster and revised-brief groups still look like product key visuals. Required headline, subtitle and prices are missing as a complete set across all four variations in each repeat. For this tested setup, plan to typeset the ad separately rather than treating four outputs as four ready-to-publish posters.
- The separate reference revision checks a different question: can an existing layout survive a price and lighting change? Inspect it below rather than confusing a new full-brief generation with an edit. Midjourney Remix is quoted at 90 credits per request here, compared with 60 for a new Fast/Standard generation.
Compare the outputs
Same brief, visible differences.
Compare the outputs
GPT Image 2.6 Flare Low
3:4 · 1K · low
Prompt, settings and references
Design a vertical advertising poster for fictional NOMA citrus drink: one cobalt-blue glass bottle on pale stone, one orange slice, cool evening light. Exact headline NOMA CITRUS, exact subtitle SUMMER EDITION, exact price $10, exact offer 2 FOR $18. Clear hierarchy and generous margins. No other text.
References: 0 · high
GPT Image 2.6 Flare Medium
3:4 · 1K · medium
Prompt, settings and references
Design a vertical advertising poster for fictional NOMA citrus drink: one cobalt-blue glass bottle on pale stone, one orange slice, cool evening light. Exact headline NOMA CITRUS, exact subtitle SUMMER EDITION, exact price $10, exact offer 2 FOR $18. Clear hierarchy and generous margins. No other text.
References: 0 · high
Midjourney V8.2
3:4 · standard · standard
Prompt, settings and references
Design a vertical advertising poster for fictional NOMA citrus drink: one cobalt-blue glass bottle on pale stone, one orange slice, cool evening light. Exact headline NOMA CITRUS, exact subtitle SUMMER EDITION, exact price $10, exact offer 2 FOR $18. Clear hierarchy and generous margins. No other text.
References: 0 · —
The first repeat opens by default. Switch repeats to inspect variation; all returned images remain available. A failed task is retained in the public record and is not counted as a successful image. Loads the tested prompt, references and parameters into the generator. Review the current quote and press Generate yourself.
Revise the approved poster
A separate one-step edit starts from each model’s first poster output. Midjourney uses Remix, image 1 chosen by index before inspection; all four returned variations are shown. GPT edits the corresponding poster as a reference. This is a workflow comparison, not identical editing algorithms.
GPT Image 2.6 Flare Low
3:4 · 1K · low
Prompt, settings and references
Keep the approved poster composition, same cobalt bottle, orange slice, pale stone, NOMA CITRUS and SUMMER EDITION. Change only the price to $10, the offer to 2 FOR $18, and lighting from warm afternoon to cool evening. Preserve all other design details. No extra text.
References: 1 · high
Starting reference ↗GPT Image 2.6 Flare Medium
3:4 · 1K · medium
Prompt, settings and references
Keep the approved poster composition, same cobalt bottle, orange slice, pale stone, NOMA CITRUS and SUMMER EDITION. Change only the price to $10, the offer to 2 FOR $18, and lighting from warm afternoon to cool evening. Preserve all other design details. No extra text.
References: 1 · high
Starting reference ↗Midjourney V8.2
3:4 · standard · standard
Prompt, settings and references
Keep the approved poster composition, same cobalt bottle, orange slice, pale stone, NOMA CITRUS and SUMMER EDITION. Change only the price to $10, the offer to 2 FOR $18, and lighting from warm afternoon to cool evening. Preserve all other design details. No extra text.
References: 1 · —
Starting reference ↗The first repeat opens by default. Switch repeats to inspect variation; all returned images remain available. A failed task is retained in the public record and is not counted as a successful image. Loads the tested prompt, references and parameters into the generator. Review the current quote and press Generate yourself.
Waiting once is not finishing the job.
Read the complete chain alongside its quoted cost. A fast image that breaks the label may need a correction; an expensive image may still miss the brief. We have not priced hypothetical retries or human retouching.
| Tested setup | Requests | Observed wait · Median | Range | Credits per three requests |
|---|---|---|---|---|
| GPT Image 2.6 Flare LowCompare the outputs | 9 | 32.3 s | 25.8–33.1 s | 210 |
| GPT Image 2.6 Flare MediumCompare the outputs | 9 | 28.0 s | 26.4–37.6 s | 450 |
| Midjourney V8.2Compare the outputs | 9 | 217.8 s | 214.0–230.9 s | 180 |
Midjourney returns four variations per request; GPT was requested at one. We report request totals and all outputs separately. Dividing by four is not the cost of four usable ads.
Observed waits include submission, reference transfer, queue, network and two-second polling. Up to three requests were in flight. They describe this batch, not guaranteed model speed.
These are CreateVision AI credit quotes for the tested settings, not API prices or a universal price per image. Check the current quote before generating.
How we tested
We fixed three full text-to-image briefs before generating and ran each three times per configuration. Each request is independent: stage three changes the complete brief, not an earlier output. Midjourney V8.2 ran Fast/Standard and returned four variations; Flare ran Low or Medium at 1K with High planning depth and one requested image. We publish every returned variation. No output was retouched or rerolled for quality. We then ran a separate one-step reference revision on the first poster output of each repeat. GPT used an image edit; Midjourney used Remix with its first variation, chosen by index before inspecting quality.
Visual assessments use a fixed checklist: requested change completed, protected details preserved, required words correct, no unwanted additions. They are editorial judgments, not biometric identity scores.
The identical brief is a reproducible baseline, not the best prompt possible for every model. We did not run a separate model-optimized prompt arm or a budget-matched quality sweep. Our conclusions apply to these settings and tasks.
CreateVision AI sells access to the compared models. This is an in-house editorial test, not an independent laboratory ranking. We publish the originals and our criteria so you can disagree with the assessment.
No blind preference panel was conducted. We describe composition rather than claiming a measured aesthetic winner.
A small set of fictional subjects cannot establish reliability for every face, product, language or campaign.
Official capabilities and tested availability
An untested feature does not count as a win. GIF, SVG, markup and web search are outside the shared image-task comparison. Official GPT Image documentation and the live model pages are linked below.
Questions before you choose
Does the revised full brief prove editing consistency?
No. The third full-brief stage starts again from text. A separately labeled reference revision starts with each model’s first poster output. Midjourney uses Remix and GPT uses an image edit; these are different production workflows, disclosed rather than treated as identical algorithms.
Which model makes more beautiful images?
We did not run an anonymous preference panel, so there is no measured aesthetic winner here. Compare the originals for your brand; the checklist assesses brief compliance separately.
Why show all four Midjourney variations?
Choosing only the best one would hide the real selection workload. All four returned images are public, and the quote is shown per request rather than presented as four approved advertisements.
Try the task with your own image.
Choose an output above to load its exact setup. Replace the fictional reference with your own and check the quote before creating.
