I was trying to find out if the results will be better if I replace the words by their functions in the sentence, like replacing “I would like to go outside now” by “pronoun verb verb preposition verb adverb adverb”.
I was trying it in different ways for several months and my results were either same as with words or even worse 🙁
At the end, after two years o so, I completely abandoned the idea of improving the results in this way, but then I tried to use it for calculating the results, I thought that it would be useless, too, and it worked!
Probably generating realistic faces using only speech as input was the hardest for me, mostly because nobody had done that before. This was a project that lasted for about 8 months, and had to ask many different people to get involved in the project in order to help me. We finally managed to do it, we named the project “wav2pix”, and since we published our work many people around the world have managed to improve our results, luckily for science 🙂
Comments
Francisco commented on :
Probably generating realistic faces using only speech as input was the hardest for me, mostly because nobody had done that before. This was a project that lasted for about 8 months, and had to ask many different people to get involved in the project in order to help me. We finally managed to do it, we named the project “wav2pix”, and since we published our work many people around the world have managed to improve our results, luckily for science 🙂