Evaluation of WebRTC as a Framework for Voice Recordings in Online Surveys

Authors: Anabell Hacker, Iris Sidonie Bakker, Ingo Siegert

Abstract:

Web-based speech data collection is increasingly important for large-scale linguistic and phonetic research. WebRTC enables browser-based audio recording without native apps, potentially lowering participation barriers. This study compares three WebRTC-based recording frameworks, native WebRTC, RecordRTC, and MediaStream Recorder, integrated into the online survey platform SoSci Survey. Speech data were collected under varying recording settings and across different devices, operating systems, and browsers. Recordings from 80 participants were manually analysed for technical artefacts. All frameworks showed similar artefact types, including clipping, noise, echo, and amplitude distortions, which occurred more frequently in sustained vowels than in read speech. Artefacts were most prevalent on Android and Windows systems using Chrome, while Apple devices showed fewer distortions. Avoiding enforced audio formats reduced clicking artefacts on some platforms.