---
{"title": "How to draft exhibit labels by speaking in 2026", "description": "A plain guide for museum curators to draft exhibit labels by speaking, with steps, the tools to try and the mistakes to avoid.", "url": "https://xpdaily.com/how-to-draft-exhibit-labels-by-speaking-for-museum-curators/", "date": "2026-10-28"}
---
A plain guide for museum curators to draft exhibit labels by speaking, with steps, the tools to try and the mistakes to avoid.

Draft label copy by speaking, then check it: dictate a first draft with a microphone, keep each label to its planned length, fix object names and dates, and match the museum's voice.

| Option | What it does | Fee |
|---|---|---|
| Paper notes | You draft label text by hand and edit it on the page | Not published |
| Word processor with dictation | You type by voice inside a document and revise the text | Not published |
| QuickDictate | Types what you say into whichever window you are using [^1] | Free for personal and noncommercial use [^1] |

## What you need before you start

Drafting label text by speaking takes only a few things.

- **A microphone and a working dictation setup.** Check that the microphone captures your voice clearly. Try it on a museum computer before you write a real label.
- **The label brief.** Know the object's approved name, the facts to cover, the word limit and the tone. The brief decides what you dictate.
- **A place to catch the draft.** A word processor, a note window or paper all work, as long as the spoken text can land there.
- **Notes on pronunciation.** Object titles, artist names and loan terms often need spelling fixes, so list them before you start dictating.
- **Access checks if you use a tool.** Some dictation setups bill you through a separate provider or need a model installed first. Check that before you begin.

Set aside a block of time before you draft. If you are working while the gallery is busy, check whether background noise affects recognition.

## Step by step

These steps work with any tool, including a pen and paper.

1. **Open the window you want the text in.** Pick the word processor, content form or notes app where the label will land.
2. **Check the microphone.** Do a short test line and confirm the words come through clearly.
3. **Say the object names once out loud.** This warms up the hard sounds before the real draft.
4. **Dictate a first draft from the brief.** Say the facts in plain words, in the order the label needs them. Do not polish yet.
5. **Keep punctuation simple.** Add stops and commas only where natural. Cleanup works better than dictating full punctuation.
6. **Stop and read the draft.** Read it back before changing anything, so you catch what the tool heard wrong.
7. **Fix names, dates and spellings.** Correct the object title, artist names and any numbers the tool missed.
8. **Trim to the label length.** Cut the draft down to the word limit in the brief.
9. **Match the museum's voice.** Adjust the wording so it reads like the labels the museum already publishes.
10. **Save the draft and move on.** File it where the next person in the label workflow can reach it.

Draft one label fully before doing several at once. Try one label and review the result before you start a batch.

## Tools that can do it

You do not need special software for this. Here are options to consider.

**Paper notes.** You write each label draft by hand and edit it on the page. Nothing is published here about price or limits because it depends on what you already have.

**A word processor with dictation.** Check whether your computer or word processor has a voice typing feature that lets you speak inside a document. Check its own help page for limits, cost and setup, since facts about specific apps are not provided here.

**QuickDictate.** QuickDictate is made by the team that publishes this site. QuickDictate is one option for drafting label text by speaking. It is a Windows program that types what you say into whichever window you are using. [^1]

You hold a key while you talk, or press it once to start and once to stop, and you can change both keys. [^1] What you say is typed into whichever window has focus, such as an editor, a terminal or a web form. [^1]

You can use ElevenLabs, Deepgram, OpenAI, AssemblyAI, DashScope or Google, and that provider bills you directly. [^1] Each provider is a separate service that bills you directly for its own service. [^1] Choose Local in the settings menu and install a speech model, and dictation then needs no internet connection or API key. [^1] With five of the cloud providers, words appear while you are still talking. [^1]

You can list words and phrases that are sent to the provider to help it recognise them, and specified spellings such as "Chat GPT" can be corrected automatically to "ChatGPT". [^1] Profiles can change the provider, language, vocabulary and formatting depending on which app is in focus. [^1]

You can browse and filter your last 50 dictations in the settings menu, and the history survives a restart or self-update. [^1] Saying "scratch that" discards the text you just dictated instead of inserting it; it is the only voice command shipped. [^1] While you are dictating, other apps' audio is muted or turned down, and it goes back to its old level when you stop. [^1] If you sign in, you can sync preferences such as hotkeys and language between PCs. Keys, audio and transcripts are not part of that sync. [^1]

QuickDictate is free for personal and other noncommercial use, with no subscription or account. [^1] Cloud providers bill you separately for their service. [^1] From v0.9.0 on it is source-available under PolyForm Noncommercial 1.0.0, and commercial use needs a separate licence. [^1] It runs only on 64-bit Windows 10 and 11; there is no Mac or Linux version. [^1] It is delivered as an .exe file on GitHub, with a built-in updater that works silently, and it can also be built from source. [^1]

## Common mistakes to avoid

**Leaving errors in dictated text.** Read every label back once before you pass it on and correct anything that does not match the intended wording.

**Skipping spelling checks.** Check artist names, object titles and loan terms against approved records. Keep a list of spellings that need attention and correct them in each draft.

**Skipping the word limit check.** Compare the draft with the brief and cut it to the required length before the label reaches the designer.

**Ignoring possible noise-related errors.** Check the draft for errors that may have resulted from background noise. If needed, move to a quieter space and dictate unclear lines again.

**Starting without checking the setup.** A tool may need a separate provider or a model installed first. Sort that out before you dictate a real label.

**Drafting every label at once.** Do one first, review the first draft for recognition errors and formatting, then continue with the rest.

## Frequently asked questions

### Can QuickDictate draft a whole exhibit label by speaking?

Yes. QuickDictate is a Windows program that types what you say into whichever window you are using. [^1] Saying "scratch that" discards the text you just dictated instead of inserting it, which helps when a sentence goes wrong. [^1] Read the draft and edit it before you publish.

### Does QuickDictate work offline?

It can. Choose Local in the settings menu and install a speech model, and dictation then needs no internet connection or API key. [^1] The other way is to use a cloud provider such as ElevenLabs, Deepgram, OpenAI, AssemblyAI, DashScope or Google, which bills you directly. [^1] If internet access is unavailable on the museum network, check whether the local option meets the museum's requirements.

### What does QuickDictate cost for museum use?

QuickDictate is free for personal and other noncommercial use, with no subscription or account, and cloud providers bill separately for their service. [^1] Cloud providers bill separately for their own service, so the museum pays the chosen provider directly for that service. [^1] Commercial use needs a separate licence. [^1] Check the licence terms against the museum's use before relying on it.

### Can QuickDictate fix how object names are spelt?

Yes. You can list words and phrases that are sent to the provider to help it recognise them, and specified spellings such as "Chat GPT" can be corrected automatically to "ChatGPT". [^1] Check the resulting text to confirm that artist names, loan terms and object titles are spelled correctly.

### Which computers run QuickDictate?

It runs only on 64-bit Windows 10 and 11, so there is no Mac or Linux version. [^1] Check compatibility with the museum's existing computers before choosing QuickDictate. It is delivered as an .exe file on GitHub, with a built-in updater that works silently, and it can also be built from source. [^1]

QuickDictate is free for personal and other noncommercial use, with no subscription or account. Cloud providers bill you separately for their service. [^1] - [Visit QuickDictate](/go/start)

## Sources

[^1]: QuickDictate's own description of itself (the product of this site's publisher) (/go/start)