tidyoutput.
← All tools

AI Prompt Splitter

Paste a document that is too long for one message. The splitter breaks it into numbered parts at sentence boundaries, each with its own copy button.

Long text

When you need to split a prompt

Every AI chat has a limit on how much text one message can carry, and long documents, transcripts, or codebases regularly blow past it. Pasting an overlong prompt gets it silently truncated or rejected. The reliable workaround is to feed the text in parts, and to tell the model to hold its answer until everything has arrived.

How the splitting works

  • Clean boundaries. The text is cut at paragraph breaks where possible, then at sentence ends, so no part stops mid-thought.
  • Numbered parts. Each chunk is labeled Part 1 of N, so neither you nor the model loses track.
  • Built-in instructions. With the instructions option on, every part tells the model to acknowledge and wait, and the final part asks for the full answer. Turn it off if you only want the raw pieces.
  • Size presets. The presets are conservative so the parts fit comfortably in typical message limits, leaving room for your own framing text.

A tip for better answers

Before part 1, tell the model what the text is and what you will eventually ask, for example: "I will paste a contract in 4 parts. After the last part, list the obligations for each party." Models handle long input better when they know the goal up front.

Frequently asked questions

How many characters fit in one ChatGPT message?

It varies by plan and model, and the practical limit is often smaller than the advertised context window. The Standard preset of about 15,000 characters is safe for most current chat interfaces.

Why split at sentences instead of exact character counts?

Cutting mid-sentence makes the model misread both halves. Clean boundaries preserve meaning, so each part is slightly under the limit rather than exactly at it.

What do the built-in instructions say?

Every part except the last asks the model to reply with a short acknowledgment and wait. The final part says the text is complete and asks the model to respond to it as a whole.

Does this work with Claude and Gemini too?

Yes. The parts are plain text with universal instructions, so they work in any chat interface.

Is my document uploaded?

No. Splitting happens in your browser and the text never leaves your device.

More tools