Voiceitt is the world’s leading personalized speech recognition platform for people with non-standard speech. Whether you’re just getting started or looking to improve your experience, this guide walks you through everything you need to know about how Voiceitt learns, adapts, and improves-especially when you help it help you. In this post, we’ll cover:
-
- How to train Voiceitt to understand your unique voice
-
- What to expect from training levels
-
- How Voiceitt continues learning as you speak
-
- How to self-annotate your spontaneous speech
- How to submit feedback to improve your accuracy
Step One: Train Voiceitt with Your Speech
Voiceitt starts by building a custom speech model based on your voice. To begin, you’ll record at least 50 phrase cards, using speech that reflects how you naturally speak. Once you’ve completed your first 50 recordings, Voiceitt’s system generates a basic recognizer model within two hours. You can then begin exploring Speak Mode and Dictate Mode, but remember: 50 phrases is just the beginning. The more recordings you add, the better Voiceitt will understand you.Speak Naturally
It is critical to record your phrases as close to your natural speech as possible. Try not to over-articulate or exaggerate your pronunciation. Avoid sounding like you’re reading aloud-this often differs significantly from how you speak in everyday conversation. The goal is to teach Voiceitt how you truly sound.What Are Voiceitt’s Training Levels?
Voiceitt uses a training level system to measure your progress. Each level corresponds to a fixed number of recordings and an improvement to your voice model. Model updates happen about 2–3 hours after you complete each level. That’s the best time to test your recognition accuracy in Speak or Dictate Mode. You can stop at any level if you’re happy with performance-or keep going to further improve it. There’s no maximum number of recordings, but there is a smaller return on efforts beyond 1000 recordings.Are The Phrases Too Hard?
No problem. Voiceitt offers multiple phrase set options and the ability to create your own by adding Personal Vocabulary for names, work terms, or custom phrases. You can easily switch by clicking “Change Phrases” while in Record Mode.-
- Short (1–2 words)
-
- Simple Language (2–4 words)
-
- Medium (1–4 words)
-
- Default (1–6 words)
Voiceitt Keeps Learning-Even When You Stop Recording
Voiceitt doesn’t just learn from your recordings-it keeps adapting through everyday use. When you use Speak or Dictate Mode, Voiceitt sends anonymized audio to Voiceitt’s team of professional annotators, who:-
- Listen to the audio
-
- Compare what they hear to the app’s transcription
-
- Use their corrected transcript with the audio to tune your speech model to improve future accuracy
How to Self-Annotate for Better Accuracy
While the annotation team handles ongoing tuning, you can play a direct role in improving your model by self-correcting transcripts. Here’s how to do it:-
- Open Dictate Mode in the Voiceitt app, in a quiet place without other voices present.
-
- Speak naturally to tell a story (e.g. Talk about a favorite movie or memory)
-
- When finished, correct the transcript directly in the text box
-
- Repeat as needed
-
- Copy the corrected text
-
- Email the corrections to support@voiceitt.com and CC support@thebabelgroup.com
-
- Be sure to include the date, time, and time zone of your session
-
- Email the corrections to support@voiceitt.com and CC support@thebabelgroup.com
Prefer a Video Walkthrough?
Watch this short Loom video from Dr. Rachel Levy, SLPD CCC-SLP, co-founder of The Babel Group, to see how to self-annotate step by step. You’ll learn exactly how to:-
- Open and use Dictate Mode
-
- Avoid background noise
-
- Make corrections
-
- Submit your feedback the right way