RNIB Study Report: Synthetic Speech for Audio Description
The synthetic speech for audio description study set out to evaluate the quality of synthetic voices, their suitability for AD and their impact on viewer experiences.
Working with UK broadcasters RNIB used synthetic voices to voice audio description scripts created for the study. These were then presented to participants via an online survey and as part of a focus group. To encourage an objective review it was not disclosed that they used synthetic voices until after participants had provided their views.
Key insights
Suitability of synthetic voices
- In most cases participants did not identify the voices as being synthetic,
- The voices were identified as lacking emotional impact for genres such as sports and drama but this was considered less of an issue for factual content
Preferences for audio description
- Participants wanted emotional resonance for entertainment genres such as drama and sports,
- Participants valued ‘human aspects’ of audio description such as when the describer laughed or when participants could hear amusement or excitement in their voice,
- Participants preferred when audio description matched the content in factors such as accents, language used and emotional resonance