Taking Midjourney from pilot to production means four things: choosing the right subscription for commercial use, building a shared prompt and style library, setting rules for rights and disclosure, and adding a review step before any image reaches a campaign. Teams that skip these steps end up with inconsistent visuals, unclear ownership and images that quietly break brand guidelines.
What is Midjourney good at, and what is it not?
Midjourney is an image generation model run by an independent research lab of the same name. It is accessed through its web app and, originally, through Discord. It is known for strong aesthetic quality, painterly and photographic styles, and features such as style references, character references, image prompts and in-app editing for variations, zoom and repainting regions.
It is less reliable for exact text in images, accurate product shapes, logos and technical diagrams. It does not know what your product looks like unless you give it a reference, and even then fine details drift. Treat it as a tool for mood, concept and background imagery rather than product photography.
Phase one: run a short, scoped pilot
A pilot should answer one question: can Midjourney produce usable images for a specific kind of asset faster or cheaper than your current process? Keep it narrow.
- Pick one asset type, such as blog headers, social backgrounds or campaign moodboards.
- Name two or three people who will run it for four to six weeks.
- Record time per approved image and how many generations it took.
- Collect every approved image with its full prompt and parameters.
- Note each rejection reason: off-brand colour, wrong anatomy, unwanted text, legal concern.
At the end, the rejection log is often more useful than the approved set. It tells you which rules your production guide needs.
Which plan do you need for commercial work?
Midjourney’s terms grant ownership of generated images to paying subscribers, with an extra condition for larger companies above a revenue threshold, who need a higher tier. Check the current terms on the Midjourney website, since plans and conditions change.
Privacy is the other plan question. By default, images made on lower tiers are visible in the public gallery. Higher tiers offer a private or stealth mode. If you are working on unreleased campaigns or client work under NDA, that mode is not optional.
Buy seats for named users rather than sharing one login. Shared accounts make it impossible to trace who generated what, and they break the audit trail you will need later.
Phase two: build the production kit
Once the pilot proves value, turn what you learned into shared assets the whole team can use.
A style reference library
Midjourney’s style reference feature lets you point at existing images to steer the look. Save a small set of approved references and their codes for each campaign or brand line, and store them where everyone can find them.
Prompt templates
Write templates that separate fixed brand elements from variable subject matter: “[subject], soft daylight, muted teal and sand palette, 35mm photograph, shallow depth of field”. New users then change only the subject.
Parameter defaults
Agree on aspect ratios per channel, the model version you use and stylisation settings. Mixing versions mid-campaign produces visibly different results.
How should rights and disclosure be handled?
Ownership under Midjourney’s terms does not settle every legal question. In several jurisdictions, purely AI-generated images may not qualify for copyright protection, which means competitors could reuse them. For key brand assets, such as a hero image you want to protect, consider having a designer substantially rework the output.
Never prompt with the names of living artists or photographers to copy their style for commercial work, and do not use real people’s likenesses without consent. Avoid trademarked characters and products. These are the requests most likely to create disputes.
For disclosure, follow the platform rules where you publish. Meta, TikTok and YouTube have labelling policies for realistic synthetic media, and the EU AI Act adds transparency duties for certain deepfake-style content. A simple internal rule works: label anything that could be mistaken for a real photo of a real event or person.
Phase three: the production checklist
Every Midjourney image going into a live asset should pass this review.
- Brand: colours, mood and composition match the style guide for that campaign.
- Anatomy and objects: hands, faces, text fragments and product shapes checked at full size.
- Rights: no artist names, real people or third-party trademarks in the prompt.
- Privacy: generated in private mode if the campaign is unreleased.
- Record: prompt, parameters, user and date saved to the asset library.
- Edit: final retouching done in Photoshop or Figma, with resolution upscaled for print if needed.
- Disclosure: label applied where platform policy or internal rules require it.
What does a realistic first quarter look like?
In the first month, expect most time to go into prompts and rejections rather than finished images. That is normal and is where the style library comes from. By the second month, templates should cut the number of generations per approved image, and designers start spending more time on retouching than on prompting.
By the end of the quarter, the team should know which asset types Midjourney handles reliably and which it does not. Write that list down. It prevents new colleagues from wasting hours trying to make the tool produce packshots or infographics it was never good at.
How Alien Road applies Midjourney
In a typical engagement our design team starts with a discovery session to pull the client’s brand guide, past campaign imagery and any legal restrictions. We then run a scoped pilot on one asset type and keep the rejection log described above. From that we build a style reference set and prompt templates specific to the client.
In production, designers generate concepts in private mode, select and retouch them, and attach the prompt record to each delivered file. The client approves images through the same review flow as any other creative. Monthly reporting covers images delivered, average generations per approved image and time spent in retouching. For example, a client’s blog might receive consistent header images every week from the same template, while product photography stays with a photographer.
Where to start this week
Choose one asset type you produce every week and give two people a four-week Midjourney pilot on a commercial plan with private mode. Ask them to log every prompt, approval and rejection reason. At the end, turn the rejection reasons into your first production checklist. That document, more than the images, is what makes the tool safe to scale.
Keep reading: Graphic Design · AI Marketing