Imagen Prompt Builder

How to Write a Great Imagen Prompt

Imagen rewards photographic terminology — focal length, lighting direction, depth-of-field cues, and specific style references. Treating Imagen prompts like camera settings produces more consistent, professional results.

The builder above structures your prompt in the order Imagen handles best: subject, setting, lens, lighting, style, then details. Copy and paste into Playcut or any Imagen-powered surface.

The six slots Imagen actually pays attention to

Imagen 3 and the newer Imagen 4 family parse prompts in roughly this order: subject and framing first, environment second, lighting third, lens and depth-of-field fourth, style and reference cues fifth, and post-production details sixth. When prompts get reordered or when a slot is skipped, the model substitutes its statistical average for that slot — and statistical averages produce stock-photo aesthetics. Here's what each slot actually controls and the words that move the needle:

• Subject and framing. Name the subject in concrete terms — gender, approximate age, hair color, wardrobe, expression, body posture. Then state the framing: close-up portrait, medium shot, full-length, environmental. Imagen reads framing words as compositional instructions, not as decoration.
• Environment. Where, when, and with what grounding props. "Golden-hour light on a Brooklyn rooftop, the BQE in the distance" is specific. "Outside" is not. Imagen renders environment depth and detail in proportion to how specific the prompt is.
• Lighting direction and quality. Pick a source ("window light from camera left," "softbox key with rim from behind," "available daylight") and a quality ("soft and diffused," "hard with deep shadows"). This is the single biggest lever for making AI images look photographed rather than rendered.
• Lens and depth of field. Focal length (24mm wide, 50mm natural, 85mm portrait), aperture-equivalent depth ("shallow depth, subject in focus, background falling off"), and any optical character ("subtle barrel distortion at the edges," "gentle vignette"). Imagen knows what these phrases mean and renders accordingly.
• Style and references. Color register, film stock, and photographer references. "Editorial commercial style, muted earth tones, Kodak Portra 400 grain, in the style of Annie Leibovitz" stacks four signals that compound. Imagen weighs named-photographer references heavily — pick photographers whose work matches your target aesthetic.
• Post-production details. Final retouching cues — "skin retouching at the level of a commercial campaign, no overexposure, preserved film grain in shadows." This slot prevents the over-smoothed plastic look most generic Imagen prompts produce.

Imagen Prompt Builder FAQ

How does Imagen compare to Midjourney or Flux for product work?

Imagen is the strongest single model for photographic realism on people and products when the prompt names specific lighting and lens vocabulary. Midjourney is stronger for stylized, illustrative, or fashion-editorial work where the brief is less literal — the free Midjourney prompt generator builds the comma-keyword flag format it expects. Flux is strong for text-on-image and complex compositional control — the free Flux prompt generator writes the natural-language paragraph prompts it favors.

When the brief moves from stills to video, the free JSON prompt generator builds the validated JSON prompt objects Veo, Sora, and Kling parse.

The honest answer is that good art directors use all three — Playcut routes between them automatically based on the prompt's intent, but if you're using Imagen directly, this builder gets you the most out of it. If a draft still feels flat, the free AI prompt enhancer layers in lighting, lens, and mood detail before you generate.

Can I use Imagen output for social link previews?

Yes — a 1200×630 export works as an Open Graph card. If you just need the share card itself (title, logo, brand colors) without generating a photo, the free OG image generator draws and exports one directly in your browser.

Does Imagen support reference images?

Yes — Imagen 4 supports a "reference image" input alongside the prompt. This builder generates the text portion of the prompt; for reference-image workflows, paste the text output into Playcut's reference-to-image flow and upload the anchor image alongside it. Using a reference is the most reliable way to keep a custom AI actor's face consistent across scenes.

Why are my generations producing the wrong race/age/body type?

Imagen has documented biases when subject descriptors are vague. Specify all four of: approximate age, ethnicity, body type, and wardrobe. Vague prompts get filled in with the model's statistical center, which usually doesn't match what you wanted. The builder above prompts you for these explicitly to prevent the drift.

How do I get a logo or text rendered correctly?

Don't render text in Imagen if you have an alternative. Generate the photo cleanly, then composite logos and text in a separate step (Figma, Photoshop, or Playcut's on-product compositing flow). Even Imagen 4 still produces unreliable text — the only way to ship a real campaign with on-image text is to add it in post.

Want the full studio?

Playcut routes Imagen and Gemini for images, plus Veo for video — all from one chat. Start your 7-day trial — pick a plan, card required, cancel anytime.