I Built a Tool to Fix My Bad Midjourney Prompts

A lot of AI prompt tools try to do everything.

They help with ChatGPT, image generation, copywriting, brainstorming, coding, and whatever else can fit into the landing page. That sounds broad and useful, but in practice it often means the output becomes generic.

That was the problem I kept seeing with Midjourney prompting.

If you already use Midjourney regularly, you probably know the pattern:

• the idea in your head is clear enough
• the prompt you type is rough
• the result comes back close, but not really right
• then you start stacking more adjectives, style words, and random references
• and the prompt gets longer without getting much better

So instead of building another general-purpose AI prompt helper, I built a small tool focused on one thing:

turning rough image ideas into cleaner Midjourney prompts with more useful settings.

That tool is Midjourney Prompt Optimizer.

Why I focused on Midjourney only

I think a lot of prompt tools fail because they are built from the tool-maker’s perspective, not from the actual workflow.

From the tool-maker side, it is tempting to say:

one prompt box, one AI model, many use cases

From the user side, that usually creates weak output.

Midjourney is not just “another AI text box.” It responds to a particular mix of:

• subject clarity
• composition hints
• visual hierarchy
• lighting and mood direction
• parameter choices like aspect ratio, stylize, chaos, and weird

So if the system treats Midjourney like generic prompt writing, the output often becomes bloated or vague.

I wanted something narrower and more practical.

Instead of trying to generate a “perfect” cinematic super-prompt every time, the goal is to produce a better first draft:

• clearer subject
• cleaner structure
• stronger framing direction
• more usable settings
• less random keyword stacking

The real problem is not prompt length

One thing I noticed early is that many people assume weak Midjourney output comes from prompts being too short.

I don’t think that is usually the real problem.

The real problem is more often one of these:

1. the subject is unclear
2. the image goal is mixed or contradictory
3. the style words are doing too much work
4. the settings do not match the intended result
5. the prompt is adding noise instead of direction

That is why I designed the product around structure instead of prompt inflation.

For example, a rough input like:

dark fantasy knight

should not automatically become a giant wall of decorative terms.

A better transformation is something more like:

dark fantasy character portrait of a battle-worn knight in blackened steel armor, standing in a ruined temple, cold blue rim lighting, drifting ash and fog, dramatic cinematic framing, detailed metal textures, grounded heroic stance

Then the tool can suggest settings that actually fit the use case, instead of leaving the user to guess.

What the product does right now

The current version focuses on a simple workflow.

You start with a rough concept. Then the tool helps generate:

• a cleaner Midjourney-ready prompt
• a suggested aspect ratio
• stylize guidance
• chaos guidance
• weird guidance when appropriate

That means the output is not just "better wording." It is meant to be a more usable starting point for actual image generation.

Right now the product is aimed at:

• creators
• designers
• marketers
• beginners who want better first drafts

The free plan gives a few optimizations per day, and paid plans are there for people who use Midjourney more seriously in ongoing creative work.

What I learned while building it

A few things became clearer once I started shaping this into a real product instead of just a landing page.

1. Generic AI UX is easy to ship but weak as a product

It is very easy to build a text area, a submit button, and some nice marketing copy.

It is much harder to make the output consistently feel tailored to a real use case.

For this product, the real work is not "make the UI look like an AI startup." The real work is:

• understanding intent better
• classifying the image goal correctly
• resolving conflicting prompt directions
• recommending settings that feel sensible
• making outputs feel less templated

2. Before/after examples matter more than abstract claims

People do not care much when a tool says "get better prompts."

They care when they can see:

• what the rough input was
• what changed
• why the optimized version is more usable

That is one reason I think example pages and use-case pages matter so much for a product like this.

3. The first draft is the real product promise

I do not think tools like this should pretend to eliminate iteration.

Creative work still needs taste and refinement.

But getting from a weak rough idea to a strong first draft faster is already valuable.

That is the promise I want to keep the product honest about.

Why I'm writing this now

The site just launched, and I'm now in that awkward early phase where the product is real enough to use, but still early enough that outside feedback matters a lot.

I'm especially interested in feedback from people who:

• use Midjourney often
• build creative tools
• have strong opinions about prompt quality
• think most AI products are too generic

If that sounds like you, I'd love for you to take a look.

Try it

The product is here:

mjpromptoptimizer

If you try it, I'd love to know:

• whether the prompt output feels more useful than generic prompt helpers
• whether the settings suggestions feel sensible
• which use cases should be improved first
• where the output still feels too templated

Early products usually do not need more hype. They need sharper feedback.

That's what I'm after now.