Forum Discussion
Neural Voices - Specific words mispronounced / errors with generated closed captions
I've noticed when implementing additional spacing (I use a paragraph from notes as my narration) that I tend to lose random words or whole sentences in the closed caption. It creates a lot of QA QC time to ensure everything is where it should be / closed caption edits / additions.
The neural voices are getting better, but I always have a word or two that come up by reviewers as sounding strange, then I spend a considerable amount of time trying to spell it differently / make it sound like what I wrote, right now for example, the word I'm struggling with is "data".
Any tips for overcoming these issues? Perhaps I should explore the SSML feature?
Hi Sarah,
I commiserate with: "I tend to lose random words or whole sentences in the closed caption". I had that problem too, extremely time consuming. I found no solution, bar laboriously typing (copy/pasting) the missing words into the captions. In cases of mispronunciation I rewrote the sentences (where possible) if it was taking too long to come up with a phonetic spelling that was pronounced correctly.
I have used a number of neural / AI voice generators that clients had paid quite a bit for, and to date they all have a pronunciation problem and tonal problems. Clients went back to using real people/voice-over artists. Also, AI/neural seem to be averse to short words with two syllables.
For data - dahder or day-ter worked for me.Hi Sarah, have you tried changing the accent (if that is permitted) to British one? Emma can speak almost all words correctly.
Hello Ange,
Thank you so much for your input, and I feel your pain! Let's hope for a solution here and I'll try your suggestion for the word DATA. I had tried so many versions, but not these two.
Hello, I haven't tried that yet. I have hundreds and hundreds of slides with narration... Is there a way to change all sounds clips at once? Although then I have to add all my spacing back in and line up many objects as it seems every narrator has different pauses / cadence.
Hi sarahrydgren-ab!
Happy to jump in here!
Updating text-to-speech globally across a Storyline project isn't supported yet, but we are tracking a feature request to give authors that ability. I've included your voice in the feature report being tracked and will update this discussion if it makes it onto our Feature Roadmap.
Have a great start to your week!
Hi StevenBenassi please add my name to the request list for this feature.
Thanks
