VisualPrompts
en
Open the builder
Free · no sign-up · 5 image + 8 video models

Direct the shot.
We write the prompt.

Pick framing, scene, light and lens like a photographer would. VisualPrompts turns those choices into a detailed prompt — formatted the way Midjourney, ChatGPT, Nano Banana, Flux, Veo, Sora, Kling or Runway read best.

Start building — freeBrowse examples
01 — Your choicestry it
Framing
Scene
Lighting
Camera
02 — Formatted for
upper-body portrait, woman in a burgundy evening dress, dark upscale evening lounge, low-key warm amber light, ~2800K, phone front camera, 24mm, extreme photorealism --ar 4:5 --style raw --v 7 --no text, watermark, CGI
Short phrases, key ideas first, parameters at the end.

Two ways in.

Pick options, or just describe the shot in a sentence.

01Prompt builderPick around 90 options across 15 sections and copy the prompt for any model.02Text to promptDescribe the shot in a sentence, in any language, and get a full brief.

One set of choices.
Thirteen dialects.

Every model reads prompts differently. Same selections, re-formatted for each with one click.

01MidjourneyImageShort, concrete phrases with the most important ideas first. Aspect ratio and exclusions go in parameters at the end.02ChatGPTImageFollows long, specific instructions. Structured JSON keeps scene, subject, wardrobe, light and camera separate.03Nano BananaImageGemini’s image model prefers descriptive sentences — a scene briefed the way a photographer would.04FluxImageReads natural language closely but has no negative prompt, so everything is phrased positively.05Qwen-ImageImageHandles long detailed prompts and a separate negative prompt to steer away from artifacts.06VeoVideoEach part of the shot on its own line: who, action, where, camera move, lighting and audio.07SoraVideoReads like a screenplay shot: scene in prose, then cinematography, then actions and dialogue.08KlingVideoCompact prompts focused on subject and motion, followed by scene, camera and lighting.09RunwayVideoMotion-first prompts: the camera move, then the scene, then the details — with no negative prompt.10LumaVideoOne flowing, descriptive paragraph with the camera motion spelled out.11SeedanceVideoSubject and action first, then scene, camera and style on their own lines.12HailuoVideoNatural description plus bracketed camera commands such as [Push in] or [Tracking shot].13WanVideoDetailed natural language — subject, scene, motion and look — with a separate negative prompt.

Three takes. That's it.

TakeScene 01
01

Start anywhere

Load a template, describe the shot in a sentence, or paste a prompt you found online.

TakeScene 02
02

Fine-tune by section

Framing, scene, subject, wardrobe, lighting, camera, color and style — about 90 simple choices across 15 sections.

TakeScene 03
03

Copy for your model

Pick the target and copy. JSON or text, negatives and parameters are formatted for you.

For people who'd rather not write prompts.

01

Motion-first video

For video models the action leads: beats in order, natural secondary motion and camera movement — not a frozen photo description.

02

Photographer vocabulary

Lens, angle, depth of field, key-light direction and color temperature — without knowing the terms.

03

JSON or text

Structured JSON for ChatGPT and Gemini, comma phrases for Midjourney, labeled lines for Veo.

04

Polish with AI

Optionally let Claude rewrite the draft for your model, using your own Claude.ai plan.

05

Consistent by design

Conflicting choices are resolved as you pick, so a prompt never asks for two things at once.

06

Private

No account. Your selections and history stay in your browser.

Start from a shot.

All examples →
Night glam selfieImageQwen-Image
Night glam selfieDark upscale interior, amber light, burgundy dress
Golden hour street portraitImageQwen-Image
Golden hour street portraitSunset, 85mm, creamy bokeh
Studio product shotImageQwen-Image
Studio product shotClean backdrop, softbox, ad quality
Fashion editorialImageQwen-Image
Fashion editorialHard flash, magazine-cover feel
Neon cyberpunk nightImageQwen-Image
Neon cyberpunk nightRainy street, neon reflections
Night selfie clipVideoWan 2.2
Night selfie clipA 5-second clip of the night glam scene
Neon street walkVideoWan 2.2
Neon street walkTracking shot through a rainy neon street
Product reveal clipVideoWan 2.2
Product reveal clipSlow orbit around a product on a wet pedestal

Questions.

What is VisualPrompts?+

A free prompt generator for AI image and video models. You pick options — shot size, location, subject, wardrobe, lighting, camera, composition, color and style — and it writes the prompt in the format your model expects.

Which models are supported?+

Image: ChatGPT (GPT Image), Nano Banana (Gemini), Qwen-Image, Midjourney and Flux. Video: Veo, Sora, Kling, Runway, Luma, Seedance, Hailuo and Wan. The same selections re-format for any of them with one click.

Is it free? Do I need an account?+

It’s free and there is no sign-up. Your selections and history are stored only in your browser.

Does it work for video?+

Yes. Pick Veo, Sora, Kling, Runway, Luma, Seedance, Hailuo or Wan and the prompt is written the way video models read it: the action first, then camera, light and sound.

Can I just describe what I want?+

Yes. Write a sentence in any language and it becomes a full, detailed specification you can fine-tune and copy for any model.

Why do prompts differ between models?+

Midjourney prefers short phrases and parameters, Flux and Runway have no negative prompt, ChatGPT handles long structured JSON, and video models need separate lines for action, camera movement and audio.