Free AI Prompt Enhancer
Turn a rough image idea into a clearer prompt. Choose a descriptive paragraph or comma-separated booru tags, and guide the AI with your own instructions. Everything runs locally in your browser, with no account or tokens needed.
Your directions guide the enhancement. The AI will still return only the prompt in your chosen format.
Device requirements & first download
Recommended: a computer with 8 GB RAM, about 2 GB GPU memory and 2 GB free browser storage. A working WebGPU browser is required. Phones may not have enough memory.
The first download is roughly 1 GB and may take several minutes. The compatibility model can be larger. Your browser caches the model when storage is available. Clearing site data, private browsing or browser cleanup can require another download.
Keep this tab open while it works.
Your result stays available.
Write a prompt, choose the format, then press Enhance Prompt.
Make image prompts more specific
A useful image prompt communicates the subject, pose, setting, composition and lighting. This tool uses a small language model running on your GPU to expand or organize that idea, while asking it to preserve your original subject and intent.
For a natural-language model, a woman in a garden could become an adult woman standing among garden flowers, framed at eye level in warm afternoon light. A tag-based model may work better with solo, garden, flowers, standing, sunlight. These are format examples, not a guaranteed output or image quality improvement.
- Write or paste the prompt you want to improve.
- Choose Natural language or Booru tags. Optionally enable Instruct the AI to ask for a specific change.
- Press Enhance Prompt. Review and edit the result, then copy it or use it in the image generator.
If you are starting from an image, the local image captioner can create booru tags first. Send those tags here to build a new prompt from the visible image features.
Prompt enhancer questions
Does my prompt leave my device?
No. The language model runs locally in a browser worker on your GPU. The browser downloads the model and runtime, but your input prompt, extra instructions and result are not uploaded. They are saved in this tab's browser storage to survive the page reload.
Should I choose natural language or booru tags?
Choose natural language for a concise descriptive paragraph, for example an adult red-haired woman standing in soft sunlight. Choose booru tags for short English labels separated by commas, for example red hair, solo, standing, sunlight. Use the format preferred by your image model.
What does Instruct the AI do?
It lets you add directions such as focus on lighting, keep it short, or preserve the outfit. These directions are appended to the enhancer's built-in instructions. The AI should still return only the enhanced prompt in your chosen format.
What device and browser do I need?
A computer with 8 GB RAM, about 2 GB GPU memory and 2 GB free browser storage is recommended. A working WebGPU browser is required. The first download is roughly 1 GB and can be larger for the compatibility model. The page chooses a full-precision variant when half-precision shaders are unavailable. Phones may not have enough memory.
Will the AI always preserve every detail?
The enhancer is instructed to preserve your subject, names, intent and model or style tokens, but a small language model can still make mistakes or decline requests. Review the editable result before using it. More detail does not always produce a better image.
Continue your image workflow
Generate an image from your prompt, describe an existing image, upscale an image, edit with layers and brushes, or remove a background.