Midjourney direction desk
Upload a reference and turn its visible art direction into a concise Midjourney prompt. The result describes the subject, medium, setting, light, color, mood, and composition, then keeps parameters separate so the instruction is easy to test and edit.
Minimal product photograph of a white wristwatch, overhead composition, pale gray studio surface, brushed metal reflections, diffused softbox light, quiet commercial mood
Frame
--ar 4:5
Variation
--chaos 8
Style pressure
--stylize 150
Exclude
--no text, logo
Midjourney input
Generate the six outputs, open the Midjourney version, and check that the visual phrase ends before --ar, --no, stylize, or chaos controls begin.
No image selected
Add an image to create all six prompt formats.General · Midjourney · FLUX · Stable Diffusion · ChatGPT Image · Nano Banana
What belongs in the text
A useful instruction names what should appear and gives the model a clear visual direction without narrating every decision. Midjourney documentation recommends short, simple phrases and asks users to think about subject, medium, environment, lighting, color, mood, and composition. That makes image analysis a good starting point: the reference already contains those choices, but they must be compressed into language the model can scan quickly.
This generator does not claim to recover the hidden text that made an image. Pixels do not store the original prompt, model version, seed, edits, or generation parameters. The tool reads visible evidence and writes a new prompt that aims at the same creative territory. Treat it as reverse art direction. You are rebuilding the brief, not extracting a secret caption.
The distinction protects your expectations. The generated text can carry a centered camera angle, soft window light, coral and cobalt color contrast, editorial styling, and a restrained mood into a new generation. It cannot lock every face, object edge, word, or spatial measurement. If exact identity or composition matters, combine the text with the platform reference controls rather than forcing more adjectives into it.
One analysis on this site creates General, Midjourney, FLUX, Stable Diffusion, ChatGPT Image, and Nano Banana versions. The model-specific version is written independently for Midjourney instead of being a renamed General prompt. That matters because a portable description can be complete yet too literal, too long, or too instruction-heavy for the way Midjourney interprets creative phrases.
Prompt order
The best sequence is practical rather than ceremonial. Put the identity of the picture first, move through the choices that shape it, and keep parameters at the end. This makes a Midjourney prompt easier to diagnose after the first generation.
Start the Midjourney prompt with the person, object, place, or event that must survive every variation. Name pose or action when it changes the silhouette. “Adult cyclist holding a road bicycle” carries more useful structure than “adventurous traveler,” because the model can place the body and object before it decorates the scene.
Say whether the target is a studio photograph, watercolor, 3D render, editorial illustration, ink drawing, or another medium. A Midjourney prompt benefits from a specific treatment because medium affects texture, edge behavior, color handling, and what kinds of detail feel believable.
Place the subject in a readable setting and state the relationship that controls scale. “Small figure on a stone bridge facing a crystal portal” gives the Midjourney prompt a foreground, destination, and size contrast. A loose list of “bridge, portal, clouds, fantasy” leaves those relationships undecided.
Describe the light source and its behavior before adding mood words. Warm rim light through cool mist explains why a scene feels cinematic. The Midjourney prompt should connect mood to visible causes such as limited color, soft transitions, hard shadow geometry, or broad atmospheric depth.
Finish the descriptive phrase with framing, viewpoint, and focal emphasis. Add parameters after the prompt text, following Midjourney syntax. Aspect ratio changes the canvas, stylize changes how strongly the model applies its aesthetic training, chaos widens variation, and the no parameter can discourage unwanted elements.
Parameters after description
A Midjourney prompt has two jobs. The text describes what the image should be; parameters control how the generation behaves. Mixing the two makes revisions harder. “Vertical composition” can remain in the creative direction, while “--ar 4:5” defines the canvas. Put every parameter at the end with the required spaces and dashes.
Use only parameters you understand. A reference image may suggest a portrait canvas, but it cannot tell you the original chaos or stylize value. Those values are creative choices for the next run. The generated Midjourney prompt can propose a sensible starting point, yet you should change one control at a time and compare the grid before adding another.
Midjourney also has Image Prompts, Style References, and other reference features. Text and reference inputs solve different problems. The Midjourney prompt explains subject and intent in words; a visual reference can carry color, content, composition, or style more directly. When similarity matters, use both and state which qualities should remain stable.
Describe is another useful comparison. Midjourney says its Describe feature returns prompt suggestions and will not precisely copy the uploaded image. This site follows the same honest boundary while adding five other platform versions and a structured visual analysis. The goal is a controllable starting point, not a guarantee of identical pixels.
Grid testing
A strong first draft still needs a controlled test. Midjourney produces variation by design, so judge the prompt by which relationships survive across the grid, not by whether one image happens to match every detail.
Compare the Midjourney prompt with the reference before copying it. Check the primary subject, viewpoint, light direction, dominant color relationship, material finish, and mood. Correct a wrong camera angle before polishing minor props. Early words carry the visual identity of the request.
Start with the Midjourney prompt and a necessary aspect ratio. Avoid stacking experimental parameters on the first pass. The baseline tells you whether the descriptive phrase works. If the composition drifts, rewrite the relationship instead of trying to fix it with style pressure.
Keep the Midjourney prompt stable while changing stylize, chaos, or the reference strength. Then keep the setting and light but replace the subject. One-variable tests show which part of the instruction the model followed and prevent random improvement from being mistaken for a reliable method.
Signed-in users can edit and save generated text in History. Keep the version that produced a useful relationship, even if the final image needed another pass. A short note about the working ratio or reference role is more useful than a folder full of nearly identical drafts.
Drift diagnosis
Most misses come from unclear priorities, not a shortage of adjectives. Diagnose the first broken relationship and revise that part of the Midjourney prompt.
A Midjourney prompt that opens with atmosphere, quality labels, and style references may let the model decide the subject too freely. Move the concrete subject and action to the front, then place mood after the visual facts that create it.
Long clauses such as “make sure that” and “please preserve” consume space without adding imagery. Rewrite the instruction as direct visual phrases. Save conversational constraints for tools that are designed around dialogue.
A wide aspect ratio can destroy the portrait balance you intended to study. High variation can scatter a carefully centered product layout. The Midjourney prompt and its parameters should support the same experiment instead of pulling in different directions.
Text alone will not preserve a specific person, logo, tiny inscription, or exact object spacing. Use permitted reference controls, inspect platform rules, and expect variation. No Midjourney prompt can recreate hidden generation settings from pixels.
Parameter questions
A Midjourney prompt is the text or phrase that tells Midjourney what image to create. A useful version identifies the subject and visual treatment, then adds setting, light, color, mood, composition, and any needed parameters. It can be short. Clear priorities usually matter more than length.
No. The image does not reveal the exact Midjourney prompt, seed, version, parameter values, reference weights, or later edits. The tool analyzes visible choices and writes a new prompt that can guide a similar creative direction.
There is no required word count. Midjourney recommends short, simple prompts. Keep enough detail to define the subject and the relationships that matter, then remove phrases that only praise quality or repeat the same idea.
Include camera language when it communicates a visible result such as viewpoint, focal length feel, depth of field, motion, or studio treatment. Do not present inferred exposure settings as factual EXIF data. They are creative directions for the next image.
Parameters belong at the end of the Midjourney prompt after the descriptive text. Check the current Midjourney documentation for supported parameters and syntax, because platform controls can change between versions.
Visitors receive three daily credits. One image uses one credit and produces all six prompt versions, including the Midjourney prompt. An account adds editable History and permanent registration credits. Batch input is available to Pro subscribers.
Start with one controlled grid
Upload the reference, read the Midjourney prompt from subject through composition, and edit the first inaccurate relationship before you copy it. Use the generated text as a creative brief, add only the parameters needed for your test, and pair it with permitted reference controls when similarity matters. A concise Midjourney prompt leaves you room to see which direction actually changed the result.