🧰 UtlKit

AI Image Prompt Generator

Pick a model, describe your subject, then tap the cards to add composition, lighting, color, and style details. Your prompt builds live and copies in one click.

Prompt language: English

Prompt preview

Describe your subject or tap the cards above to start building your prompt.

🐛

Report a Problem

Found a bug or have a suggestion? Help us improve this tool.

📊 Data Summary (auto-filled)

Tool: t2i-prompt-generator · /tools/t2i-prompt-generator/

model: Seedream

language: en

selected: [object Object]

What is this tool?

The AI Image Prompt Generator builds structured prompts for text-to-image AI models. You describe your subject, then tap image cards across subject type, scene, composition, lighting, color palette, style, medium, mood, camera, and aspect ratio to stack details. It supports models like Midjourney, DALL·E, Flux, and Imagen, appends a per-model quality suffix, and lets you copy the result as text or JSON, all in your browser.

When to use this tool

  • Prompt Chinese models in Chinese

    Select Seedream, Kolors, or Wan on the Chinese page and get a ready-to-paste prompt written in Chinese, including visual terminology the model expects.

  • Build layered prompts for Western models

    Stack composition, lighting, color palette, and style cards to produce a layered English prompt for Midjourney, DALL·E, Flux, Imagen, or Stable Diffusion that covers every axis the model supports.

  • Learn visual design vocabulary

    Each card pairs a named technique with a reference image, so you can explore compositions, lighting setups, color palettes, and camera angles while building a prompt and reuse the terminology in your own descriptions.

How to use

  1. 1

    Choose the target model

    Select your target model from Seedream, Kolors, Wan, MiniMax Image, ERNIE ViLG, Midjourney, DALL·E, Flux, Imagen, or Stable Diffusion 3.5.

  2. 2

    Describe your subject

    Enter your subject, scene, or concept in free text — it becomes the first line of the prompt.

  3. 3

    Add visual details

    Tap the image cards across ten categories — subject type, scene, composition, lighting, color palette, style, medium, mood, camera, and aspect ratio — to stack details into your prompt.

  4. 4

    Copy the prompt

    Preview the prompt live, then copy it as plain text or as JSON with the full option list.

Frequently Asked Questions

Which image models are supported?

Ten models are covered: Seedream, Kolors, Wan, MiniMax Image, and ERNIE ViLG from Chinese providers, plus Midjourney, DALL·E, Flux, Imagen, and Stable Diffusion 3.5. Selecting a model appends a quality suffix tuned to that model at the end of your prompt.

How do the option categories work?

There are ten categories with 92 options in total: subject type, scene, composition, lighting, color palette, style, medium, mood, camera, and aspect ratio. Each category is a single choice: tapping a card selects it, tapping another card in the same category replaces it, and tapping the selected card again removes it. Your chosen fragments are assembled in fixed order after your subject text.

What language will my prompt be in?

For the Chinese models (Seedream, Kolors, Wan, MiniMax Image, ERNIE ViLG), using this tool on the Chinese page generates the prompt in Chinese; every other combination produces English, since these models respond best to it.

Is my data uploaded to a server?

No. The entire tool runs in your browser: the prompt is assembled locally from your text and chosen options, and nothing is sent to any server. The reference images are static files served with the page, and your subject text stays on your device.

Related tools