How do I use pronunciation in Synthesia?

📌 Pronunciation Controls are currently being rolled out gradually. If you don't see this interface yet, you will soon.

The pronunciation feature lets you control how specific words are spoken in your videos. You can adjust pronunciations for acronyms, technical terms, brand names, and other terms to ensure your video sounds natural.

Set a pronunciation

Highlight the word in the script editor.
Select Pronunciation.
Choose how to input your pronunciation:
- Type — enter a phonetic spelling manually.
- Record — speak the word aloud and the system will transcribe it into a phonetic spelling.
The pronunciation will automatically be previewed so you can hear how it sounds.
If it isn't right, select Back and adjust your input and preview again.
Select Use to save the pronunciation. You can then:
- Select Apply to all to apply this pronunciation to every instance of the word in your current video.
- Select Add to glossary to save it to your workspace Glossary, so it applies automatically across all videos in your workspace.

✍️ Words with a saved pronunciation appear italicized in the script editor.

📌 Add to glossary is only available on Enterprise plans. To learn more, check out How do I use the Glossary in Synthesia?

Voice Persistence

Our Voice Engine now stores generated audio permanently at the paragraph level for each voice. This means that as long as the text in a paragraph stays the same, the audio for that paragraph will remain consistent — whether you’re previewing, generating, or regenerating a video.

This ensures your voice output stays stable and predictable across edits, so you can make changes confidently without unexpected variations in tone or delivery.

✍️ This only applies for videos generated from mid-November 2025 onward.

Speech Regeneration

If you want to manually update audio, use the Regenerate button next to a paragraph. This will generate new audio for the entire paragraph.

💡 To undo this change, click Undo or press Ctrl+Z / Cmd+Z.

Audio is also automatically regenerated for any paragraph where you edit the script— previous audio will not persist after script changes to a paragraph.

Pronunciation Examples

Acronyms

✍️ Acronyms can be tricky—some are pronounced as full words (like NASA), while others are read letter by letter (like W-H-O). If the avatar doesn't pronounce an acronym the way you expect, you can use the pronunciation tool to spell it phonetically and ensure it sounds just right.

Example	Solution
*Acronym as word:* ERP/NBA is pronounced as one word instead of spelling it out.	*Acronym as word:* Use the pronunciation function and write 'e r p' or ‘en be ay’. Use spaces and small letters.
*Acronym as individual letters:* STAMP/ASAP is pronounced like S-T-A-M-P/A-S-A-P but you actually want it to be pronounced as one word.	*Acronym as individual letters:* Use the pronunciation function and write 'stamp' or ‘eighsap’. Use small letters.

Pronouncing foreign words

✍️ In some cases, when you’re using words from different languages in your script, you may need to amend the pronunciation.

Example	Solution
Writing a script in English and using the French word ‘rendez-vous’.	Use the pronunciation function and input ‘rahn-day-voo.’

Phonetic Spelling Table for Foreign Words

Letter	Phonetic Spelling	Alternatives
A	ai	ah, ay, eigh
B	bee	be, buh
C	see	kuh, suh, cee
D	dee	de
E	ee	eh, eei
F	ef	fuh
G	jee	guh
H	eitch	huh, (silent)
I	eye	aiy, ah-ee
J	jay	jai
K	kay	kai
L	el	ell
M	em	em
N	en	nuh
O	oh	ah, oh, uh, oo
P	pee	puh
Q	cue	kwuh
R	ah-r	arr, are
S	ess	zuh
T	tee	tuh, tea
U	yoo	yew
V	vee	vea, vuh
W	double-you	double-yew, duboyoo
X	ecks	exx
Y	wah-ee	wye, wyie
Z	zee	zea, zuh

Homographs

✍️ A homograph is one spelling shared by two different meanings, sometimes with different pronunciations

Example	Solution
The miners used a special tool to *lead* the way, but were unsure if it would detect lead in the sample.	Use the pronunciation function to adjust pronunciation. In this case, lead (the metal) is mispronounce.

Pronouncing long words

✍️ Sometimes AI voices struggle with pronouncing very long words. Especially in other languages than English, this issue is quite common.

Example	Solution
Some voices might struggle with a long word like *‘antidisestablishmentarianism’*	Use the pronunciation function and split long words into single words. In this case, ‘anti-dis-establish-men-tar-ian-ism.’

Numbers and Symbols

✍️ When necessary, you can spell out numbers and symbols using full words:

573 ➡️ five hundred seventy-three or five seven three
12:30 p.m. ➡️ 12 thirty p m or twelve thirty in the afternoon
29 Willow Street ➡️ twenty-nine Willow Street
support@companyname.com ➡️ support at company name dot com

💬 FAQs

Can I write pronunciation for non Latin alphabet?

The pronunciation feature supports all alphabets. Using the alphabet of the targeted language would, however, yield the best results.

For instance, if you use a Japanese voice, you could use a Latin alphabet for the pronunciation but leveraging hiragana is better.

How do I set pronunciation across a workspace?

Any workspace user can add terms to the glossary by default. If you are a workspace admin, you can restrict this so that only admins can add or edit entries — go to Workspace settings and toggle Restricted glossary management. To learn more, check out How do I use the Glossary in Synthesia?

Is the voice speed change always exact?

Voice speed values are approximate. For example, setting the speed to 1.2x won’t always make the voice exactly 20% faster — the final pacing depends on your script and how the AI voice interprets it.

📚 To learn more, check out the Synthesia Academy Feature Fridays: Pronunciation recording.

How do I use the user settings in Synthesia?

How do I create a Studio Express-1 Avatar?

How do I use the Glossary to manage translations and pronunciations?