Text-to-speech voice data
Studio-quality recordings for building synthetic voices, read by professional voice artists in the language and accent you need. Our network of voice artists around the world records on demand, and every take is checked against the script before it reaches your dataset.
- Script matchEvery word read as written
- LevelsNo clipping, steady volume
- NoiseClean room tone, no mouth clicks
- Pronunciation"7:30 p.m." read as "seven thirty p m"
0412.wav48 kHz, 24-bit, with transcript- Voices
- Professional voice artists around the world, recording on demand
- Recordings
- Single-speaker voice builds, multi-speaker sets and expressive styles
- Audio
- To your spec, for example 48 kHz, 24-bit WAV
- Output
- Checked audio files, matching transcripts and session notes
What TTS voice data is
Text-to-speech (TTS) voice data is a set of clean studio recordings of a speaker reading a prepared script, paired with exact transcripts. A TTS model learns the voice, accent and speaking style from these recordings, so any noise, misreading or change in tone ends up in the synthetic voice.
That's why the script and the checking matter as much as the voice. We write scripts that cover the sounds of each language and check every line before it goes into your dataset.
What we record
In the languages, accents and styles your product needs.
Single-speaker voice builds
Many hours from one voice artist, recorded over several sessions with a consistent sound.
Multi-speaker sets
Shorter recordings from many voices, balanced by gender, age and accent to your quotas.
Expressive styles
The same lines read in different styles, such as cheerful, calm, apologetic or newsreader.
Script writing
Phonetically balanced scripts in each language, plus your product's own names, numbers and phrases.
Line-by-line QA
Every file checked for misreadings, noise, clipping and level, with retakes recorded where needed.
Pronunciation notes
How names, abbreviations and numbers should be read, agreed before recording.
How a voice data project runs
You choose the voice from auditions before recording starts.
Spec
Languages, accents, voice profile, styles, hours and audio format.
Auditions
Short samples from shortlisted voice artists for you to choose from.
Script and terms
Scripts finalized and usage terms agreed with each artist.
Recording
Sessions recorded in professional studio quality, on a schedule.
Line QA
Every line checked against the script and spec, with retakes as needed.
Line check QA
Related AI audio services
Questions about TTS voice data
What is TTS training data?
TTS training data is a set of high-quality recordings of a speaker reading a script, with transcripts that match the audio exactly. Text-to-speech models use it to learn to generate speech in that voice.
How many hours of recordings does a TTS voice need?
It depends on the model and how expressive the voice must be. Tell us your model's requirements and we'll plan the script and number of sessions to match.
Where do the voice artists record?
Our voice artists around the world record on demand in professional studio quality, to the audio spec you set. We check every file before delivery.
Do the voice artists agree to AI use of their voice?
Yes. Usage terms, including use for training and synthetic voices, are agreed with each artist before recording starts, and the terms are shared with you.
Request voice samples
Tell us the languages, accents and voice profile you need, or book a call. We send auditions from shortlisted voice artists.
- Phone
- +44 20 3290 0116
- Phone hours
- Weekdays 9:30 to 16:00, weekends 9:30 to 12:00, UK time