← ブログ一覧へ戻る

ChatGPT・Claudeで台本を書いてVoiceScrollで読む方法

目次
  1. Short answer
  2. Ask for a script that is meant to be spoken
  3. Clean up before you paste
  4. Paste it into VoiceScroll
  5. Use square brackets for directions
  6. Read it
  7. Where your script goes

ChatGPT and Claude are good at turning an idea or a page of notes into a first draft. VoiceScroll handles the next step: reading that draft out loud while the text follows your voice. There is no direct connection between them. You copy the answer and paste it into VoiceScroll, which takes less than a minute. This guide covers what to ask for, what to clean up before you paste, and how to mark directions so they don't trip you up.

Short answer

  1. In ChatGPT or Claude, ask for a script written to be spoken, in plain text, with any directions in square brackets.
  2. Copy the answer.
  3. Open VoiceScroll and paste it into the script box.
  4. Press start and read. The script moves as you speak and waits when you pause.

Ask for a script that is meant to be spoken

A script you read aloud is different from one you read silently. Ask for short sentences and everyday words. Ask for no headings, bullet points, or emoji, because they show up as symbols and aren't words you would say. Say how long it should run when read aloud, who it is for, and what tone you want.

Here are three requests you can copy and change. Replace the words in capitals with your own.

  • From an idea: "Write a script I will read aloud on camera for about 60 seconds, introducing TOPIC to AUDIENCE. Use short sentences and plain words. No headings, bullet points, or emoji. Put any stage directions in square brackets, like [pause]. Plain text only."
  • From notes or slides: "Turn these notes into a script I can read aloud in about 3 minutes. Keep my points in the same order, write it the way I would say it, and put a pause direction in square brackets between sections. Plain text only. NOTES"
  • From a script you already have: "Rewrite this so it is easy to read aloud. Split long sentences, write numbers and symbols the way I would say them, and keep the meaning and the length. Plain text only. SCRIPT"

VoiceScroll's free tier covers scripts up to 300 characters. That is a very short script, so the examples above will go over it. To try the free tier first, ask for "under 300 characters" in place of the length. Pro lifts that limit.

ChatGPT conversation: the request for a 60-second script and the start of the answer, with [pause] in square brackets
The first request above in ChatGPT, with the topic and audience filled in. The answer puts directions in square brackets.
Claude conversation: the same request and an answer that opens with [smile at camera]
The same request in Claude. VoiceScroll is not affiliated with OpenAI or Anthropic.

Clean up before you paste

  • Leave out the AI's opening line, such as "Sure, here's your script," and any offer at the end to make changes.
  • Remove formatting symbols that came along, such as # before a heading or ** around bold text.
  • ChatGPT's copy button can put a backslash in front of brackets, as in \[pause\]. Delete the backslashes.
  • Read it out loud once. If a sentence makes you stumble, ask for a simpler version of just that sentence.

Paste it into VoiceScroll

Copy the answer in ChatGPT or Claude, then open VoiceScroll. On iPhone and iPad, an empty script shows a Paste button while you have text copied. On Android, tap the box that says "Paste your script here" and paste. On Mac, click the script box and press Command-V. The Mac app runs on Macs with Apple silicon (M1 or later) and macOS 26 or later.

VoiceScroll home on iPhone: an empty script with the Paste button
On iPhone, an empty script shows a Paste button while you have text copied.

Use square brackets for directions

VoiceScroll treats words in square brackets as script directions. While you read, it doesn't wait for you to say them, and it takes them out of the line and shows them on their own as you reach them. One use is stage directions such as [smile] or [pause]. Another is a name that speech recognition tends to mishear: you still say the name, but the script doesn't stop there.

Square brackets are the default. You can switch to parentheses or braces, or turn directions off, under Script Directions in the app's settings.

VoiceScroll reading screen on iPhone: “smile at camera” shown on its own over the script’s first lines
Claude’s script in VoiceScroll. The direction it opens with, [smile at camera], is taken out of the script and shows on its own.

Read it

Press start and read at your own pace. VoiceScroll listens and moves the script to match, so pausing to think or repeating a line doesn't make you lose your place. If you would rather not read aloud word for word, choose Auto-scroll in Scroll mode and the script moves at the speed you set.

Where your script goes

Whatever you type into ChatGPT or Claude, including your notes and the draft, goes to OpenAI or Anthropic and is handled under their own policies. If a script is confidential, check what that app does with your conversations before you paste it in.

Once you paste the script into VoiceScroll, the app stores it on your device, and we never receive it. On Android it may be included in the phone's own cloud backup. By default, speech recognition runs on your device too. On iPhone and iPad with iOS or iPadOS 27 and Apple Intelligence, VoiceScroll's own Script AI sends the script to Apple's servers only when you ask it to.

使ってみませんか?

音声認識で台本が自動で追従します。貼り付けて、開始して、自然に話すだけ — 手を触れる必要はありません。

VoiceScroll を無料で試す

話すペースに合わせてスクロールする音声連動テレプロンプター。

関連記事

テレプロンプターを使ってもロボットっぽく聞こえない方法

テレプロンプターを使っていても、文字を読んでいるように聞こえてしまう人へ向けた7つのコツをまとめました。視線の動かし方、一文の長さや区切り方、話す速さ、リハーサルの進め方を少し整えるだけで、台本を読んでいる印象をぐっと減らし、聞き手の目を見て話しているように感じさせる、自分の言葉として自然に伝えられる話し方を無理なく身につけられます。

台本を読んでいるように見せずにカメラの前で話す方法

カメラの前で台本を読んでいても自然に見える6つのテクニックを紹介します。目線の高さに合わせたテキスト表示、短い視線移動、話す速さにスクロールを合わせてレンズから目を離さない工夫まで具体的に解説し、台本を読んでいる印象を与えずに、本番でもあがらず、自分の言葉として落ち着いて話すためのコツを、はじめての方にもわかりやすく整理しました。

音声連動テレプロンプターの仕組み: VoiceScrollはどう動くのか

音声連動テレプロンプターが内部でどのように動いているのかを、専門用語をかみくだいて詳しく解説します。オンデバイスで完結する音声認識、話した単語とのマッチング処理、スクロール位置を補正する仕組みを順に取り上げながら、固定速度スクロールよりもスピーチや動画撮影で自然に使える理由を、具体的な場面を挙げながらわかりやすく紹介します。

YouTube向けテレプロンプター: 撮り直しを減らして動画品質を上げる

テレプロンプターを使ってYouTube動画の撮影時間を大きく減らす方法を解説します。カメラと画面の配置、読みやすい台本の作り方、視線を保つコツに加え、撮り直しが増えてしまいがちな失敗パターンとその避け方も紹介しながら、自然な話し方で効率よく収録し、視聴者に伝わりやすく動画の完成度を高めるための実践ポイントを整理しました。