The gap between a prompt that returns garbage and a prompt that returns something you would actually use is almost never about creativity. It is about structure. Once you know the handful of elements that make a prompt work, you can write them for anything, image, text, video, audio, in about thirty seconds. This guide gives you that structure and shows you ten concrete before and after examples so you can see the pattern.
I run this framework daily across Midjourney, ChatGPT, Claude, Veo, and FLUX. It is the same seven elements every time. Learn these once and you stop guessing forever.
Why Most Prompts Fail
Most people write prompts the way they ask friends for favors. Short, casual, assuming context the AI does not have. "Make me a logo for my coffee shop" is a reasonable thing to say to a human designer who can ask follow-up questions. It is a terrible thing to say to an AI, because the AI will not ask. It will guess, and its guesses will be generic.
The fix is not writing longer prompts. It is writing prompts with the right ingredients in the right order. Length comes from completeness, not padding.
The 7-Element Framework
Every effective prompt, whether for image, text, or video generation, contains some combination of these seven elements. Not every element appears in every prompt, but the more of them you include, the more control you have over the output.
1. Subject
The literal thing the AI is making. A woman reading a book. A corporate email about layoffs. A thirty second ad for a running shoe. Be specific. "A dog" is a subject. "A black Labrador with gray on the muzzle, sitting on a wooden porch" is a usable subject.
2. Style
How the output should look or sound. For images: photorealistic, oil painting, pixel art, anime, 3D render. For text: formal, conversational, academic, witty, deadpan. For video: documentary, music video, cinematic, home-video. Style tells the AI which neighborhood of its training data to pull from.
3. Mood or Tone
The emotional register. Peaceful. Tense. Melancholy. Energetic. Hopeful. Grim. This is independent of style. You can have a photorealistic image that is peaceful or tense depending on how the subject is lit and composed. Same with writing. Mood is the single biggest lever most beginners never pull.
4. Lighting (Images and Video) or Voice (Text)
For visual outputs: golden hour, blue hour, overcast, hard studio light, candlelight, neon, rim light, backlit. Lighting carries half the visual weight of any image. For text outputs the equivalent is voice: first person, third person, omniscient, insider, outsider, teacher, peer.
5. Composition or Structure
For images: close-up, wide shot, bird's eye view, over the shoulder, rule of thirds, symmetrical. For video: tracking shot, static, handheld, drone, dolly in. For text: a numbered list, a narrative, a question and answer, three short paragraphs. Composition decides how the viewer or reader enters the piece.
6. Quality or Precision Modifiers
For images: 8K, highly detailed, sharp focus, masterpiece, award winning, shot on 85mm. These used to matter more than they do now because modern models already optimize for quality, but they still nudge the output. For text: specify word count, reading level, format (markdown, JSON, bullets).
7. Format and Constraints
What the output literally needs to be. 16:9 aspect ratio. Square format. Three paragraphs max. Must mention these five keywords. No corporate jargon. Must rhyme ABAB. Constraints are where you prevent the AI from solving the wrong problem.
10 Before and After Examples
Image Example 1: Portrait
Before: "a woman in a garden"
After: "Portrait of a woman in her thirties with dark curly hair, standing in a wild English cottage garden at late afternoon, soft diffused golden hour lighting from camera left, shallow depth of field with background bokeh, warm color palette, candid editorial photography style, shot on 85mm lens f/1.4, peaceful contemplative mood, 3:4 aspect ratio"
Elements present: subject, style, mood, lighting, composition, quality, format. The second version removes almost all guesswork. Try both in the image generator and the difference is immediate.
Image Example 2: Product Shot
Before: "a watch"
After: "Luxury men's wristwatch with a navy blue leather strap and silver bezel, floating against a dark gradient charcoal background, dramatic single key light from upper right casting a soft shadow to the lower left, commercial product photography, hyper-detailed on dial and strap stitching, minimalist composition with the watch centered and tilted 30 degrees, 1:1 square format"
Image Example 3: Cinematic Scene
Before: "detective in an alley"
After: "A trench-coated detective standing in a rain-slicked alley at 2 AM, backlit by a single flickering sodium streetlamp casting long shadows down the wet cobblestones, cinematic film noir, melancholy and tense mood, low angle wide shot showing the full figure with reflections in puddles, moody atmosphere with visible steam from a nearby grate, shot on 35mm anamorphic lens, teal and amber color grading, 2.35:1 aspect ratio"
If you want twenty more of these, the cinematic prompts collection has ready-to-copy entries by genre.
Image Example 4: Anime
Before: "anime warrior"
After: "Anime warrior girl in silver armor standing on a cliff edge overlooking an endless cherry blossom forest at dawn, Studio Ghibli inspired illustration style, hopeful and determined mood, soft morning rim light catching the armor edges, three-quarter medium shot with wind blowing cape and hair to the right, highly detailed clean line art, pastel color palette with warm pinks and golds, 16:9 widescreen format"
Text Example 1: LinkedIn Post
Before: "Write a LinkedIn post about getting promoted"
After: "Write a LinkedIn post announcing my promotion to Senior Engineering Manager. Voice: first person, grateful but not saccharine, slightly self-deprecating. Mood: proud but humble. Structure: open with a one-sentence hook, three short paragraphs on the journey, close with a specific thank you to one mentor by name (Priya Shah). Length: 180 words max. Must mention that I joined as a junior four years ago. No corporate buzzwords like 'journey,' 'excited,' 'thrilled,' or 'incredible team.'"
Text Example 2: Cold Email
Before: "Write a sales email"
After: "Write a cold outreach email to a VP of Marketing at a mid-size SaaS company (200-500 employees). Subject line under 50 characters. Body under 100 words. Voice: direct, warm, zero sales language. Open with a specific observation about their recent product launch (placeholder bracket for me to fill in). Pitch: our tool reduces their LinkedIn ad costs by 30 to 40 percent. One concrete proof point. End with a single yes-or-no CTA for a 15 minute call. Format: plain text, no markdown."
Text Example 3: Blog Outline
Before: "Outline a blog post about remote work"
After: "Outline a 1500 word blog post titled 'The Hidden Costs of Remote Work Nobody Talks About.' Audience: engineering managers who have been fully remote for three or more years. Voice: honest, slightly contrarian, backed by lived experience not stats. Structure: five sections, each with a one-line thesis and three bullet points. Opening hook idea in the first section. Avoid any discussion of productivity, Zoom fatigue, or the usual tired topics. Focus on things like career visibility, informal mentorship loss, and local community erosion."
Video Example 1: Social Reel
Before: "a video of a coffee shop"
After: "Eight-second cinematic shot inside a specialty coffee shop at 7 AM, warm morning light streaming through large front windows, steam rising from a latte on a reclaimed wood counter, slow dolly-in from medium shot to close-up of the latte art rosetta, shallow depth of field with the barista slightly out of focus in the background, documentary feel, cozy inviting mood, natural ambient audio of espresso machine hiss and quiet jazz, 9:16 vertical format for Instagram Reels"
For more vertical video prompts, see the AI video to shorts tool.
Video Example 2: Product Demo
Before: "a product video"
After: "Six-second product video of a stainless steel water bottle rotating 360 degrees against a clean minimal cream background, soft overhead studio lighting with gentle rim light, static locked-off camera with the bottle on an invisible turntable, commercial product video style, premium and clean mood, hyper-sharp focus on logo and texture, no music needed, 1:1 square format"
Build these in the video generator.
Video Example 3: Narrative Short
Before: "a short film about a lost dog"
After: "Eight-second emotional opening shot of a lost golden retriever walking alone down an empty suburban street at dusk, warm orange sky with purple shadows, cinematic handheld camera following from behind at dog eye level, melancholy and hopeful mood, slow pace, soft focus on the dog with the street falling into bokeh, ambient audio of distant traffic and wind, 16:9 widescreen"
The Order That Works
For images: subject, style, lighting, mood, composition, quality, format. In roughly that order.
For text: task, audience, voice, structure, constraints. In roughly that order.
For video: subject and action, duration, camera, lighting, mood, audio, format. In roughly that order.
You do not need every element every time. For a quick image you might only need subject, style, and lighting. For a cold email you might only need task, voice, and constraints. The framework is a checklist, not a straightjacket. Scan it, pick what matters for this prompt, write.
Four Habits That Compound
First, save every prompt that works. Not every variant, just the winners. Over a year this becomes your personal prompt library and the compound effect is enormous.
Second, change one element at a time. If you got a decent result and want to iterate, change only the lighting, or only the style, or only the mood. Changing three things at once teaches you nothing.
Third, read prompts you did not write. Browse public libraries. Note the elements they emphasize. See how other people solve problems you have not thought of yet. Starting from prompts that are working right now is faster than starting from scratch.
Fourth, learn one tool deeply before you learn five tools shallowly. Pick one model, Midjourney or ChatGPT or FLUX, and run two hundred prompts through it. You will develop intuition that transfers to every other tool.
The seven-element framework is not a secret. It is just discipline. Apply it to the next ten prompts you write and the before-and-after gap in your output will be obvious to anyone who sees your work.
