Or… Just because it sounds too good to be true doesn’t mean it is!
In March of 2015, I was contacted by one of the premiere Text To Speech technology companies, Edinburgh
Scotland-based CereProc, on behalf of a yet-to-be-disclosed US company to audition as the voice a new device and software application that runs on a very sophisticated Text to Voice (TTV) engine. After listening to 100’s (?) of voices on voice123, I was one of 12 that they invited to audition. And, after going through the audition process, I was selected.
The work was super interesting…
I spent three hours each day in studio with just me and one of two engineers (Graham and Francesca). The system we used was similar to a teleprompter. The top sentence on the screen was black, with several more upcoming sentences to be read below that were grayed out. To record, a sentence moved into top position on the screen, a low tone went off in my headphones, and then I read the sentence in an “upbeat manner” that was not “overly expressive or too flat.” When I finished the sentence, a higher tone went off in my headphones to indicate that the recording was complete. If I read it perfectly (always the goal!), that sentence disappeared, then the next sentence moved up and turned black and the process continued.
Sounds boring – and I was really afraid before I left that it would be really tough to get through the “work” part of the trip because of the nature of the job – but it was not boring at all! It was really fun and interesting. The sentences were sometimes nonsensical – or were about things that I had no familiarity with, so they didn’t make sense to me – so it required a ton of concentration. And, because each recorded sentence had to be perfect, it also required the same level of concentration from the engineers – to listen for both accuracy and fluency in reading the script, as well as clear and proper pronunciation, no (even minor) inappropriate pauses or any other minor noises that might creep in. So, we were both working really hard together – which made it fun!
When I did make a mistake – which definitely happened on some of the stuff that I wasn’t familiar with – I stopped, the “end record” tone sounded, and then the “begin record” tone sounded and we “Carry on!” We often stopped to joke about a sentence – there were a lot of “inappropriate” words and strangely uncomfortable sentences because it’s a TTV engine and you never know when someone is going to use one of “those” words…
At the end of the project working 3 hours a day, 5 days a week for 3 weeks (with about 20 minutes of break max), we went through about 10,000 sentences give or take. And, because we actually went at a very rapid pace compared to other talent they’ve had in studio, they were able to tack on more sentences to the planned script which increased the available variables and will in the end make for an even more authentic human voice sound.
This project was unusual because the US company really wants their voice to “sound real,” rather than sounding like a computer simulated TTS system. So, while most CerProc projects are two weeks long and cover a broad array of diphones, the result is nice but not quite “natural.” By adding an extra week – 1/3 more variables – and giving me the thumbs up to be more upbeat and expressive than most of the talent they work with, the goal was to have a very natural voice in the end product.
What did I learn from this?
Wow! Too much to highlight here, but in a nutshell, I learned three important things that I’ll carry with me as my career progresses:
- If an opportunity that sounds crazy knocks, don’t be afraid to check it out! Even though it sounds too good to be true… It might NOT be!
- Just because you have a tough day (as in reading the script feels impossible) don’t worry, it will be better tomorrow.
- Simply relaxing and slowing down the slightest bit will make for a much better, more accurate and articulate read (THAT’s the big one for me… since I tend to read fast a lot and have often been coached to slow down, but didn’t know how.)
What is the end product?
I don’t know yet… Don’t know who the client company is (although I have a few clues). But, if the product launches in time for the 2015 holiday season, as planned, I will know and for sure will share with anyone who’s interested.
So, there ya have it!
It was an experience and opportunity of a lifetime to be able to go to Scotland – all expenses paid – and be able to work for a great wage, get better at what I do, and explore a new city/country! After all, there were 21 hours each day that I wasn’t working!