Rita Faleiro posting from ProZ.com at 9:55 AM on 12 Nov 2025
MOS for AI generated TTS: for this job, I had to evaluate around 300 audios, which were generated by AI through TTS. MOS was used to refine and improve the quality of the voice. Some of the evaluated aspects were: mispronunciation, omission or addition of words, wrong reading of numbers/symbols/units/abbreviations, inappropriate pause/rhythm, speech speed error, noise in the audio, or even errors in the source text that could cause unintelligible or confusing parts in the audio.
