NeuralWriter

AI Prompt Generator

Turn a rough idea into a clear, ready-to-paste prompt for ChatGPT, Claude, Gemini, coding assistants, image and video AI ✎ Free, no sign-up, 30 languages.

No Sign-up Required Absolutely FREE 30+ Languages
For
Model
Tone
Detail
Language
0 / 10,000

Generated prompt

By clicking the button above, you automatically agree to our Privacy Policy and Terms of Use

Example

One idea, three prompts

These are real prompts this generator wrote for the same one-line idea, with the Concise detail level. The layout changes with the destination, and the highlighted words are placeholders for facts only you know.

Idea

Instagram post announcing our new oat milk latte

Chat · ChatGPT

Write a short, engaging Instagram post announcing our new oat milk latte to our [target audience]. Highlight the drink's appeal, create excitement about trying it, and use a warm, friendly tone that fits our brand, [brand name]. Include a clear invitation to visit or order, a few relevant emojis, and 3–5 relevant hashtags. Keep the copy natural, concise, and ready to publish. Avoid inventing product details, prices, or availability. Format: one Instagram caption of 50–80 words, with hashtags at the end.

Chat · Claude

<task> Write an Instagram post announcing our new oat milk latte to attract customers and encourage them to try it. </task> <context> Our café or brand is [brand name]. The latte's key details are [product details]. </context> <constraints> Use a friendly, inviting tone that feels natural on Instagram. Highlight the drink's appeal without inventing ingredients, flavors, health benefits, prices, or availability details. Keep the wording concise and engaging, and include a light call to action. </constraints> <output_format> Provide one ready-to-post Instagram caption of 40–70 words, including a suitable emoji and 3–5 relevant hashtags. Return only the caption. </output_format>

Image

An inviting oat milk latte in a clear glass cup on a cream stone countertop, silky foam art, scattered oats, soft morning sunlight, sage green and warm beige palette, overhead editorial café photography, minimalist Instagram launch announcement aesthetic.

Modes

Every setting, explained

Choose where the prompt goes and, for chat and code, which model will read it. The content rules stay the same everywhere: the task comes first, requirements are checkable, and nothing about you is made up. Only the layout and the details each tool needs change.

Where the prompt goes

Chat

For answers, writing, planning and analysis in a chat assistant. You get the task, who it is for, checkable requirements and the answer format, and nothing more than the request needs.

Code

For Cursor, Claude Code, Copilot and other coding assistants. The prompt names the context the assistant needs, what must not change and how to check the result, and asks for a plan before any edit.

Image

For Midjourney, FLUX, DALL-E, Stable Diffusion and similar tools. One flowing description of the subject, setting, camera, light, colours, style and mood, with no quality buzzwords.

Video

For Veo, Sora, Kling, Runway and similar tools. One continuous shot, or up to 3 for a bigger idea, with the motion and the camera movement spelled out and kept physically plausible.

Which model reads it

Any

A clean prompt in plain sentences and light structure that works in every chat assistant. The best choice when you switch between tools or share the prompt with others.

ChatGPT

Markdown headings or labelled lines where structure helps. For reasoning tasks the prompt gives the goal, the constraints and a template of the answer instead of step-by-step thinking instructions.

Claude

Each part wrapped in XML tags such as task, context, constraints and output format, as Anthropic recommends, with a tagged slot where you paste your own text or data.

Gemini

Short Markdown sections or a numbered list, direct and concise, with the main instruction first and the answer format last.

Ask first, refine after

Ask me first

Up to 3 quick questions about your idea before the prompt is written, each with ready answers to pick. Your answers go into the prompt as facts.

Shorter

Cuts repetition and anything the model would do anyway, while every requirement and placeholder that changes the result stays.

More specific

Turns vague wording into numbers, limits and clear lists of what to include and leave out, without inventing facts.

Add examples

Adds one or two short, clearly marked examples of the answer you expect, so the model sees its shape and level.

What never goes into your prompt

Most generators pad every prompt with the same tricks. They make prompts longer, not answers better, so this one leaves them out.

Fake expert role-play

No "world-class expert with 20 years of experience". A role is added only when a specific expertise changes the answer.

Invented facts

Names, numbers, dates and your audience are never made up. Anything only you know becomes a highlighted [placeholder].

Quality buzzwords

No "8k, octane render, masterpiece" in image and video prompts. Those words make pictures look plastic.

Shouting and threats

No ALL-CAPS orders, threats or tips for the model. Calm, plain instructions work better in every modern model.

How it works
  1. 01 Original Text
  2. 02 Generate Prompt
  3. 03 Generated prompt
Overview

Why use our AI Prompt Generator?

Most prompt generators turn every idea into the same long template that opens with "Act as a world-class expert". Modern models don't need that, and the padding pulls the answer away from what you asked. This tool does the opposite: a quick request gets a short, precise prompt, and only a big task gets sections.
Vague wishes become checkable requirements, such as length, number of items and what to leave out, and anything only you know appears as a [placeholder] instead of an invented fact. Choose where the prompt goes (Chat, Code, Image or Video) and which model will read it (ChatGPT, Claude, Gemini or Any), and the prompt is laid out the way that tool reads best.
Turn on Ask me first to answer up to 3 quick questions before the prompt is written, then make the result Shorter, More specific or Add examples in one click. The generator is free, needs no sign-up and works in 30 languages.
At a glance
Registration
No account needed
Price
Completely free
Models
ChatGPT, Claude, Gemini
Languages
30 supported
Prompt types
Chat, code, image, video
Character limit
10,000 per run
Clarifying questions
Up to 3, optional
One-click refine
Shorter, specific, examples
Invented facts
None, placeholders instead
Background

What makes a good AI prompt?

A strong prompt is clear rather than long. It does five things:

  • Task first. Says what to produce, for whom and why, in plain words.
  • Checkable requirements. Length, number of items and what to include or leave out, instead of "make it engaging".
  • Real facts, not guesses. Names, numbers and your audience come from you; the rest is marked as a placeholder.
  • Output format. One line on what the answer should look like.
  • A layout the model reads well. XML tags for Claude, Markdown for ChatGPT and Gemini, a plain scene description for image and video tools.

A role such as "You are an expert" helps only when a specific expertise changes the answer. This generator adds one only then, so you get a prompt that does its job without filler.

01

Prompts sized to the task

A one-line request stays a short prompt, and only a task with several parts gets labelled sections. Concise, Balanced and Deep set how far the answer should go, never how much filler goes into the prompt. Shorter prompts are also faster to read, easier to edit and cheaper to run through an API.
02

Written for ChatGPT, Claude or Gemini

Pick the model you use and the prompt is formatted for it: XML tags for Claude, clean Markdown for ChatGPT and Gemini. Keep Any for a prompt that works everywhere. The content is the same in every version, so you can switch models without rewriting your request.
03

Prompts for coding assistants

Code mode writes for Cursor, Claude Code, Copilot and similar tools. The prompt states the goal, the context the assistant needs (language, framework, files) as placeholders when you haven't given it, what must not change and how to check the result. For changes to existing code it asks for a plan first and small steps that keep the tests passing, which prevents confident rewrites of the wrong thing.
04

Image and video prompts without filler

For Midjourney, FLUX, DALL-E, Stable Diffusion, Veo, Sora or Kling you get a vivid description of the subject, camera, light and mood. No "8k, octane render, masterpiece" keyword chains that make pictures look plastic: for a photo look the prompt names a camera, lens or film stock instead. Image and video prompts are written in English by default, the language these models follow best.
05

Asks instead of guessing

Turn on Ask me first and the generator asks up to 3 quick questions about your idea, each with ready answers to pick from, or type your own. If your idea is already clear, it skips the questions. Anything you leave out becomes a highlighted [placeholder] and is listed under the result, so the prompt never invents your facts.
06

Refine in one click, free in 30 languages

Make the prompt Shorter, More specific or Add examples without starting over, and see what each version adds and why. Write your idea in any of 30 languages and keep that language or get the prompt in English. No account and no install needed.
07

Who uses our prompt generator

  • Marketers and creators. Prompts for posts, ad copy, newsletters and content plans with the audience, length and call to action spelled out.
  • Developers. Code prompts with context, constraints and checks, ready for Cursor, Claude Code or Copilot.
  • Designers and artists. Clean image and video prompts for Midjourney, FLUX, Veo or Sora without keyword soup.
  • Students and teachers. Prompts for explanations, study plans and practice questions pitched at the right level.
  • Everyone new to AI. Ask me first shows which details make an answer better, so writing your own prompts gets easier.
FAQ

Questions, answered

It is a tool, also called a prompt maker or prompt builder, that turns a rough idea into a finished prompt you paste into an AI model. You describe what you need, and the generator returns a clear prompt with the task, the requirements and the answer format, laid out for the model you use: ChatGPT, Claude, Gemini, a coding assistant, or an image or video generator.

Yes. Choose Chat and set the model to ChatGPT, and the prompt uses the Markdown structure ChatGPT reads best. It works with every current ChatGPT model, including the reasoning ones: for those the prompt gives the goal and an answer template instead of step-by-step thinking instructions.

The content is the same, the layout differs, so one tool works as a Claude prompt generator and a Gemini prompt generator alike. Claude follows prompts organised in XML tags, such as task, context and output format, especially well, while ChatGPT and Gemini work best with short Markdown sections. Pick the model in the Model switch, or keep Any for a prompt that works in all of them.

Start with the task in one plain sentence, then add what makes the answer right for you: who it is for, how long it should be, what to include and what to leave out. Skip "act as a world-class expert" and long lists of adjectives. If you are not sure what to add, turn on Ask me first and answer the questions.

Prompt engineering is the practice of writing instructions that steer an AI model toward the answer you want. With today's models it is less about tricks and more about clarity: a precise task, real context, checkable requirements and a clear answer format. This generator applies those rules to every prompt it writes.

You can, but a plain request usually comes back as a long template with a made-up expert role and invented details. This generator is tuned against exactly that: it sizes the prompt to the task, marks unknown facts as placeholders, formats it for your model, and lets you ask questions first or refine the result in one click.

Yes. Choose Image and it works as an image prompt generator, or Video for a video prompt generator. You get a description of the scene in plain words: subject, setting, camera, light, colours, style and, for video, the motion. It works in Midjourney, FLUX, DALL-E, Stable Diffusion, Veo, Sora, Kling and similar tools. Image and video models follow English best, so the prompt switches to English by default.

It asks up to 3 short questions before the prompt is written, about things that change the result most, such as the audience, the purpose or the style. Pick a ready answer, type your own or skip. If your idea is already clear, the prompt is written right away.

They are placeholders for facts only you know, such as a name, a number or your audience. The generator marks them instead of inventing details, highlights them in the result and lists them under it. Fill them in before you paste the prompt.

They set how far the answer should go: a short, direct answer, a thorough but focused one, or a comprehensive one that also covers edge cases and trade-offs. They also cap the length of the prompt itself, so even Deep stays free of filler.

Yes. You can write in any of 30 languages, and the prompt keeps the language of your idea. Switch Language to English if you want the prompt in English, which is the default for image and video prompts.

Yes. NeuralWriter's prompt generator is free and needs no sign-up. Every mode, model and setting is available at no cost, with up to 10,000 characters per run.

Fill in any [placeholders], copy the prompt and paste it into your AI tool as your message. If it is not quite right, click Shorter, More specific or Add examples to refine it, or Generate again for a fresh version.