Skip to main content
EaseTalk
Research and evidence

What we build on, and what we do not claim.

EaseTalk started as a University College Cork computer science thesis and became a product with a published paper behind it. The design decisions come from existing literature. The outcome claims do not exist yet, because we have not earned them.

EaseTalk: An LLM Driven Speech Practice Tool for Real Life Scenarios

Published through IEEE. It sets out the technical and design framework for scenario based speech practice, and it is the origin of the product.

View the publication

Read this before the evidence below

EaseTalk is not a medical device, a diagnostic tool, or a treatment. The research on this page explains why the product is designed the way it is. It is not evidence that EaseTalk improves anybody's speech. No study on this page tested EaseTalk against a clinical outcome.

The evidence base

Four findings that shaped the product.

These are the sources our design rests on. They are about speech and language therapy, not about EaseTalk.

  • Prescribed home practice is part of what works

    Brady and the RELEASE Collaborators examined dosage, intensity, frequency, delivery, tailoring, computer support and home practice for people with aphasia. The greatest language recovery was associated with frequent, functionally tailored speech and language therapy supported by prescribed home practice, at greater intensity and duration than usual clinical services internationally.

    Brady et al. / The RELEASE Collaborators (2022), Stroke 53(3), 956 to 967. Reference 1.

  • Home practice should be functionally tailored

    The Aphasia Rehabilitation Best Practice Statements recommend that functionally tailored and relevant home practice should be provided.

    Aphasia Rehabilitation Best Practice Statements (2024). Reference 2.

  • There is no single speaking challenge

    Work on individual variability in stuttering shows that fear responses, triggers and therapeutic needs differ substantially from person to person.

    Individual variability in stuttering, PMC8740569. Reference 3.

  • Rehearsal works best when it is safe and repeatable

    The literature on exposure and individual fear structures describes gradual, repeated, individually shaped rehearsal rather than one confrontation.

    Exposure therapy and individual fear structures, PMC9947508. Reference 4.

Honest limits

What we have not shown.

This section exists because a research page without one is a marketing page.

No outcome data

We have not run a controlled trial of EaseTalk, and we make no claim that using it improves speech, fluency, or intelligibility.

No diagnostic validity

Nothing in the app has been validated as a measure of anything clinical. The patterns a therapist sees describe engagement, not condition.

Recogniser bias is unquantified

We know speech recognition performs worse on atypical speech. We have not published our own measurement of how much worse, in our own product.

If you would like to change that

We welcome conversations with SLTs, clinical teams, research ethics committees and pilot partners about formal evaluation. Get in touch.

Origins and recognition

Where the work has been backed.

  • University College Cork

    Built at UCC, developed with the Speech and Language Therapy department and with people with lived speech difficulty. The research originates from a UCC computer science thesis on LLM driven speech practice.

  • Health Innovation Hub Ireland

    Winner of the HIHI AI Call 2025. Read the announcement

  • Enterprise Ireland New Frontiers

    Accepted onto the programme, and awarded the Phase II Best Pitch Award. Read the announcement

  • One to watch for 2026

    Named by Enterprise Ireland among 15 healthcare AI innovators to watch for 2026. Read the article

  • NDRC

    Partnership with NDRC.

  • South Infirmary Victoria University Hospital, Cork

    Early clinical pilot activity, bringing EaseTalk to its first clinical users.

Want to discuss clinical evaluation?

If you work in a service, a research group, or an ethics committee, we would rather hear your questions early.