Start with what you are making
A tutorial needs clear instructions. A story needs room for changes in mood. A short clip needs a strong opening. Write down the job before browsing voices.
Open Voices to explore the collection, including actors, musicians, athletes, public figures and characters. Where a recorded sample is available, use it to narrow your choices. A sample previews an existing recording; it does not read your script.
- Who is listening, and where will they hear it?
- Should the delivery feel calm, conversational, energetic or dramatic?
- Which names, numbers or words must be especially clear?
- Does this project need a familiar voice, your own voice or a new voice?
Choose a voice from the library
- Text to Speech: choose a voice from the collection and turn a script into an audio clip. This is the simplest place to start.
Test one useful paragraph
Start with one paragraph from your actual project. Include a question, an unfamiliar name and a number. Keep the words identical while comparing voices.
Generate one candidate at a time in Text to Speech. Listen for clarity and fit, then download the takes you want to keep.
- Every word is present and in the right order.
- Names and numbers sound as intended.
- The pace leaves room for the listener to follow.
- Questions and important words have the right emphasis.
- The voice suits the project without distracting from it.
Welcome to the weekend workshop. Today, we are building a small balcony garden with three pots and one bag of soil. Can it really fit in that corner? Let us measure it first.
Make one change at a time
Keep the voice and script fixed while adjusting Speed. Keep Speed fixed when trying a different voice. Otherwise, you will not know what improved the take.
- Words sound wrong: write the pronunciation or number the way you want it spoken.
- Delivery feels flat: add one supported feeling tag, such as
[calm]or[excited]. See emotion tags. - The clip runs long: shorten the script before raising Speed.
- Levels feel uneven: try Loudness normalization, then review the result. Volume adjusts the output level.
- Still not the right fit: return to the voice collection and try another candidate.
Keep your test small and save the winner
The credit cost appears before Generate. Changing the script or settings can change that cost, so check it again before each attempt. New accounts receive 50 free credits once; see Creator pricing for the current plan.
Keep the approved audio alongside the script, selected voice and settings. Generation can vary between takes. Reusing an approved intro file is more consistent than generating the same line again each week.
Project: Voice or voice description: Script version: Speed and volume: Feeling tags: Pronunciation spellings: Approved audio filename: What worked / what to change:
Frequently asked questions
Where is the model picker?
Text to Speech uses one speech experience. Choose the voice, write your script and adjust its delivery; you do not need to choose an underlying model.
Can I hear a voice before generating?
Use available recorded samples to get a sense of a voice. They are existing clips, not a preview of your new script. Generating that script uses credits.