Turn an image idea into a model-ready description and separate settings suggestions. PromptWizz writes prompts; it does not generate or test images.
Guidance checked 2026-10-20. Editorial guidance.
Editorial suggestions, not confirmed vendor limits.
0 / 10,000
Intermediate and Complex are available on Starter and Pro.
Start with the subject and the visible action. Then say where it belongs in the frame, which medium you want, and which lighting or colors matter. Quote any exact words to render, including their placement and typography. Add constraints that protect counts, relationships, or an edit you want to keep unchanged. Leave out detail that does not serve the image.
Model profiles keep the prose, tag syntax, command flags, and destination-tool settings separate. Each check date applies only to the stated evidence scope.
Describe the main subject and context in concise prose. Keep exact rendered text in double quotes. Put all command parameters in the separate parameters field. Use a style reference only when the user supplied it. Do not invent quality tags, reference codes, or deprecated reference flags.
Checked 2026-10-20: selected command parameters. Source.
V8.2 reference workflows differ from older Midjourney versions; do not use --oref or --cref here.
Use short labeled sections in this order: Scene, Subject, Details, Constraints, Use. Preserve exact text in quotes and specify typography when relevant. Keep size and quality in settings. For edits, distinguish the requested change from the elements to preserve.
Checked 2026-10-20: labeled prompt sections. Source.
Choose size and quality in the destination tool; these are suggestions, not applied settings.
Write connected narrative sentences, beginning with a clear creation or editing verb. Cover subject, action, context, composition, and style. Describe the desired positive scene. Preserve user-described reference relationships without claiming to inspect images.
Checked 2026-10-20: narrative and positive prompting. Source.
Pro is a related variant; these hints are scoped to 2.1.
Organize the description as medium, subject, setting, light, and framing. Front-load the most important visible detail. Use positive descriptions and concise concrete language. Omit generic quality tags and unsupported negative syntax.
Checked 2026-10-20: medium-first prompting. Source.
No negative prompt field or layout-coordinate generation in this workflow.
Write concise subject-first tags or short phrases, followed by setting, composition, lighting, and style. Put task-specific exclusions in the separate negative field. Avoid generic negative blobs. Do not invent weights, embeddings, LoRA names, or checkpoint-specific settings.
Checked 2026-10-20: UI-dependent separate negative field. Source.
A1111-compatible tag guidance is UI and checkpoint dependent; it is not general SD3.5 syntax.
Write a clear natural-language design brief for the Magic Prompt workflow. Put exact requested text and its typography early, followed by composition, visual style, and palette. Do not emit the open-weight JSON caption format.
Checked 2026-10-20: text-first prose and Magic Prompt. Source.
Aspect ratio is a suggestion to confirm in the destination UI; this profile does not claim 4.5 compatibility.
Write a portable visual brief covering the important subject, composition, medium, light, and constraints. Use no vendor-specific syntax or claimed model limits.
Checked 2026-10-20: editorial, not vendor-confirmed.
Editorial suggestions, not confirmed vendor limits.
Starting idea: A cozy coffee shop.
Warm editorial illustration of a small neighborhood coffee shop at dusk, two customers at the window counter, amber pendant lights, espresso machine as the focal point, open space above the window for a headline. --ar 4:3 --s 100
An editorial illustration of a neighborhood coffee shop at dusk. The espresso machine anchors the foreground; two customers sit at the window counter. Warm amber pendant lights and uncluttered space above the window leave room for later text.
Starting idea: A poster that says OPEN LATE.
Minimal street poster with the exact words "OPEN LATE" in large white condensed sans-serif lettering centered on a deep navy field; high contrast and generous margins. --ar 2:3 --raw
Design a minimal portrait poster. Render exactly "OPEN LATE" in large white condensed sans-serif type, centered on a deep navy background, with generous margins and no added copy.
These are authored examples, not generated or rendered results. The Midjourney command line follows the description.
Aspect ratio describes the frame shape. Size chooses pixel dimensions, and quality may choose a model-specific processing level. Midjourney accepts an inline --ar flag. GPT Image and other destination tools expose settings separately; a sentence in the prompt does not apply those controls for you.
The Midjourney profile adds a single --no suffix when useful. The A1111-compatible Stable Diffusion tag workflow has a separate negative field. The other launch profiles favor positive descriptions that preserve the intended exclusion without reversing it.
The image rubric checks subject clarity, composition, medium, lighting, constraints, fidelity, and model syntax. It scores the original and rewritten prompt independently. A lower or unchanged score is possible; no score predicts the image you will get.
Yes. An anonymous visitor gets one shared optimization per 24-hour window. A free account gets three; Starter gets 25 and Pro gets 1,000. Text, image, and video use the same allowance.
No. It writes image prompt text and suggests settings to choose in the image tool. You render the image there.
Yes. Choose Midjourney V8.2 for concise prose with a separate, copyable parameter line. The complete prompt puts supported flags at the end.
GPT Image is the current OpenAI image family for new work. DALL-E is historical terminology, so this tool offers a GPT Image profile.
Midjourney uses a derived --no suffix. The scoped A1111-compatible Stable Diffusion workflow uses a separate negative field. Other launch profiles use positive-first descriptions.
Usually no. Copy the image description as prompt text and select size or quality in the destination tool when that setting exists.
Describe the intended relationship to a reference here, then attach the actual image in the destination tool. PromptWizz cannot inspect or upload it.
No. The score checks how clearly the prompt specifies the task and fits the selected workflow. It is not a render-quality measure.
No. Text, image, video, and extension optimization share the same daily allowance.
Try the video prompt optimizer, text prompt optimizer, or text prompt checker. Browse templates, plans, and our PromptPerfect comparison.
Background guides: Midjourney prompting (includes older V6 advice) and prompting across models (includes historical DALL-E references). Use this page for the current launch profiles.