This is a multiple single-case study that adopts an exploratory and iterative process where the primary objective is to establish a gamified repository of high-quality audio and visual datasets (My Voice Library) of individuals 8 to 18 years with cerebral palsy and dysarthria and makes data available for research, which will be used by engineers and programmers to enable personalized real-time communication. Our secondary objectives are to (1) establish an advisory panel of people with lived experiences of cerebral palsy and dysarthria who will lend their expertise to studies using My Voice Library data as research partners; (2) develop a platform with an option to later set up a cost-effective and accessible digital health approach for dysarthria assessment and data collection; and (3) make the dataset from My Voice Library available in a program of biomedical engineering research to trial new technology innovations for real-time communication. The My Voice Library will first be piloted in 9 children with cerebral palsy, and thereafter a further 90 participants will be sought for user testing.

This study includes 1 person with lived experience of cerebral palsy and dysarthria as part of the investigator team who, together with additional parents and young people with cerebral palsy, has provided feedback on the development of the My Voice Library platform, the time and burden of completing the modules, participation information sheet, and consent forms. Additionally, our advisory panel includes people living with cerebral palsy and dysarthria, carers, and researchers in the field. Together with the My Voice Library research team, they have provided valuable ideas, feedback, and important networks. The idea for My Voice Library emerged after multiple conversations with people with lived experience of dysarthria and their challenges in finding a fast and reliable communication device that met their needs. The advisors have ensured that the format and the language of both the orientation materials and the modules themselves, as well as the study information and questionnaires included in the research development address matters important to all stakeholders. Our research team member, who has lived experience of cerebral palsy, participated in a short informational video used for recruitment and informational purposes. This video will be included in all plans to disseminate the study findings (Checklist 1).

There is no formal referral process. Caregivers of children aged 8 to 18 years who respond to a recruitment initiative will be screened for eligibility criteria (detailed in Textbox 1). Participants will be recruited through Cerebral Palsy Alliance, Australia, and any disability or consumer organizations in the national or international networks of the investigators and consumer advisors who agree to assist in circulating study information. For example, ethics-approved study information will be circulated through Cerebral Palsy Alliance’s website, X (formerly known as Twitter), Facebook, Australian CP Register, and CP Quest moderated by Cerebral Palsy Alliance.

Informed assent and consent from children and their parents/caregivers/guardians will be sought. The study participation information and consent will be embedded in a link using REDCap (Research Electronic Data Capture) [27], electronic data capture tools hosted at The University of Sydney, a secure web-based software platform designed to support data capture for research studies, providing (1) an intuitive interface for validated data capture, (2) audit trails for tracking data manipulation and export procedures, (3) automated export procedures for seamless data downloads to common statistical packages, and (4) procedures for data integration and interoperability with external sources. Additionally, within the My Voice Library account setup, consent needs will be reconfirmed.

My Voice Library participants are presented with audio-recorded adult models of each word or sentence in gamified modules based on words and sentences from the Frenchay Dysarthria Assessment—Second Edition [28] to allow speech and language therapists, engineers, and programmers to develop new innovations to enable personalized real-time communication.

It is anticipated that it will take 10 to 15 minutes to set up an account with the My Voice Library database. Part of this setup includes information about the child: geographical location, age, diagnosis, level of dysarthria (mild, moderate, or severe if a prior assessment has been conducted), communication impairment (Communication Function Classification System) [29], speech performance (Viking Speech Scale) [30], hearing and vision, language used at home, 3 questions that require the child to follow a 2-step instruction to complete, and confirmation that they have access to a laptop or personal computer with a video camera (Multimedia Appendix 1).

Before the participants start, they will be taken through an onboarding to familiarize themselves with how and what they are expected to do in the game-based modules. The parents of the children will be the main individuals to provide support; however, study personnel will be available when required.

The game-based modules are designed to create an engaging environment for children to record audio and video under the supervision of their caregiver and can be completed within 3 to 4 hours. However, this estimate is highly individual, and participants will be given the option to stop a recording and revisit it at a later stage, where they can pick up from where they left off. It is recommended that participants complete My Voice Library within 2 weeks of starting.

After completing the My Voice Library modules, study personnel will complete 2 questionnaires: the NASA TLX [25] and the SUS [26]. Each questionnaire may take about 10 minutes to complete, totaling 20 minutes. These measures will be completed with the child participant and their caregiver via telepractice/videoconferencing.

Finally, at the end of the user experience surveys, the participants and their caregivers will be invited to enter their contact details into a separate database with the view that future researchers may contact them to invite their participation in studies where they will be able to try out technology developed using audio and/or video data from My Voice Library. This section will take no more than 5 minutes to complete.

My Voice Library combines audio and videos associated with the speech sounds of children with cerebral palsy and all severity levels of dysarthria in a creative, progressive web app (Figure 1). All authors were actively involved in the design of the system, providing input and feedback throughout its development. The audio and video recordings of the children’s voices will be set up through a game-based platform (My Voice Library Database Dictionary; Multimedia Appendix 1).

Where possible, video data will be deidentified using a privacy-enabling algorithm, which prevents the face recognition software from reliably recognizing faces while maintaining facial details in the recording [31]. The following details are reported in the participant information statement provided to all participants and their families:

The children’s speech sounds will be recorded using a digital audio recorder at a 44.1-kHz sampling rate (16-bit quantization) with a microphone integrated into a computer or tablet. Video data of lipreading will be acquired using the built-in webcam of a computer or tablet. Children will hear audio-recorded adult models of each word or sentence within the game and will be asked to repeat the sounds.

The questionnaire collected information on participants’ demographic characteristics (birth year, sex, and country); primary motor type of cerebral palsy; classifications of vision, speech, hearing, gross motor, fine motor, communication, eating, and drinking; and payment information.

The SUS [26], which measures the perceived ease of use of software or hardware technology in 10 items, will be utilized. Response efficiency will be assessed using the NASA TLX [25], which measures mental, physical, and temporal demand, as well as performance, effort, and frustration across 6 items and subscales. The graphical 1 to 7 scale, with no weighted scoring, will be used. These tools will be modified for children and administered by study personnel once the My Voice Library modules have been completed. Study personnel will complete this process with children via video conferencing.

The study uses a multiple single-case design, in which each participant is analyzed individually. The sample size (n=99) was determined pragmatically, consistent with recommendations for feasibility and pilot studies, which is not intended to be powered for hypothesis testing [32,33].

The number of cases was selected to ensure sufficient representation of variability in dysarthric speech characteristics and to support the development of a comprehensive speech database. This approach aligns with current methodological guidance, emphasizing that feasibility studies should include enough cases to inform future research and methodological development. Any subgroup analyses by severity, motor type, age, or gender will be exploratory.

Descriptive statistics will be used to describe the dataset of children with cerebral palsy. Feasibility, usability, data quality, and scalability will be evaluated using a combination of quantitative and qualitative indicators. Usability will be assessed using a modified version of the SUS [26], with mean SUS scores interpreted against established benchmarks (scores ≥68 indicating satisfactory usability and ≥70‐80 indicating good usability). Cognitive workload will be evaluated using the NASA-TLX [25], with subscale and overall workload scores analyzed to identify perceived effort, mental demand, and frustration associated with using the system.

Correlations, associations, and odds ratios will be examined using appropriate statistical methods to explore relationships between participant characteristics and outcomes on the SUS and NASA-TLX. Success criteria will include feasibility, defined as achieving >70% recruitment and retention; usability, defined as an average SUS score >70; acceptable workload, defined as NASA-TLX scores within the low-to-moderate range; data quality, defined by completeness and internal consistency of collected measures; and scalability, indicated by minimal technical issues and positive user experience patterns across diverse settings.

Response accuracy, sensitivity, and specificity using both audio and video for the 99 children will produce continuous variables and will be analyzed using false positive and false negative counts to obtain sensitivity and specificity. Exploratory analysis will also be undertaken by subgroups in relation to motor type, severity, age, and gender where sufficient numbers are possible.

Exploratory supervised artificial intelligence and machine-learning approaches will be investigated using audio-only, video-only, and multimodal audio-video speech data collected through the My Voice Library platform. Potential approaches may include convolutional neural networks (CNN), recurrent neural networks (RNN), transformer-based architectures, automatic speech-recognition systems, lipreading models, and fine-tuning of existing large-scale speech-recognition models (eg, Whisper) [34]. Participant-specific speech-recognition systems may also be explored to support personalized communication approaches.

Labels for model development will be derived from the predefined target words and sentences embedded within the gamified modules based on the Frenchay Dysarthria Assessment–Second Edition, where the expected speech output is known a priori.

To minimize bias and data leakage, participant-level separation will be used between training, validation, and testing datasets. Cross-validation approaches will be applied where appropriate. Model performance will be evaluated using accuracy, sensitivity, specificity, false positive and false negative rates, and additional prediction performance metrics, as defined by the International Standards Organization [35].

As this is a protocol and feasibility-oriented research database study, analyses related to artificial intelligence and machine learning will be exploratory and intended to inform the future development of personalized speech-recognition systems rather than establish definitive clinical-performance benchmarks. Strategies to reduce overfitting will include the use of independent testing datasets, cross-validation procedures, and regularization or model-selection techniques where appropriate.

The establishment of My Voice Library is expected to enable significant advances in personalized communication technologies for children with cerebral palsy and dysarthria. By providing a high-quality, standardized repository of audio and visual speech samples, collected through gamified modules based on the Frenchay Dysarthria Assessment 2, the platform will support researchers, clinicians, and engineers to collaboratively develop and refine personalized speech-recognition models. These datasets and cross-disciplinary workflows are anticipated to improve the precision and responsiveness of real-time communication tools and ultimately enhance communicative participation for children with cerebral palsy by increasing the likelihood that their speech is understood in everyday contexts.

Building large, representative databases of dysarthric speech presents several recruitment challenges that can limit both sample diversity and data quality. The population of interest is relatively small and heterogeneous, with wide variation in etiology, severity, co-occurring motor impairments, and communication access needs, making it difficult to recruit cohorts that reflect the full spectrum of dysarthric speech [1]. Many potential participants rely on caregivers, support workers, or clinical services to facilitate involvement, introducing logistical and consent-related barriers [3]. Families may also experience research fatigue or perceive limited direct benefit, reducing willingness to participate in time-intensive protocols.

Within this context, My Voice Library offers a structured approach to mitigating some of these challenges. It embeds high-quality voice and video assessments through gamified modules based on the Frenchay Dysarthria Assessment 2 [28] and was co-designed with an active patient and a public involvement advisory group. This design enables the collection of speech data in controlled yet naturalistic environments, enhancing clinical relevance while supporting sustained engagement. The modular, gamified format reduces fatigue and increases both the quantity and diversity of data contributed by children. Additionally, the project’s information governance framework facilitates ethically approved research use of real-world data, providing families with clearer assurances around privacy, consent, and purpose.

This project describes the development of My Voice Library, a repository of high-quality audio and visual speech datasets from children with cerebral palsy and dysarthria. By enabling researchers, clinicians, and engineers to contribute to and access shared datasets, the repository holds the potential to support the development of personalized speech-recognition technologies that facilitate real-time communication. If the anticipated outcomes are achieved, My Voice Library will provide a foundation for future research and innovation aimed at improving communicative participation for children and young people with cerebral palsy who have speech impairments.