Replace "it sounded right" with criteria everyone can see, debate and agree to.
The working version of the VUX AI voice selection framework. Shortlist up to ten voices, test each one against a pass-or-fail Baseline, then score your final five on seven character criteria. Your scores save in your browser as you go and you can print the result or save it as a PDF.
Sign up to our newsletter to access the tool.
Three pass-or-fail checks: clarity, technical performance and accessibility. A voice that fails any one of them is out. There is no partial credit.
Seven criteria scored from 1 to 5: brand alignment, credibility, tone, register, naturalness, locale and accent, and demographic fit. Any score below 3 rules the voice out, whatever its total. The rest are ranked by total, out of 35.
Two or three people browse voice galleries across providers and shortlist the voices that feel about right. The obvious misfits go straight away.
Cut the long list to around ten and test each voice against the pass-or-fail checks. Fail one and the voice is out.
Score the final three to five on seven character criteria. Three people scoring independently shows you where the consensus sits.
Take the ranked shortlist to the people who sign off and make the call with the scorecard in front of you.