Please use the platform and let us know how we can improve!
It is just an alpha release. But there's some exciting offers for the first users and it is the cheapest system in the market.
Do let us know in case of any difficulty.
You can also mail us at souvic@neurals.ai for any support!
Related to uncanny valley with random name insertion, one way to combat this is to have two sentences for the greeting: (1) "Hi/Hey" and (2) <name>. This would allow designers to reduce the stress on (1) and alternate pitch and stress more freely on (2), reducing "uncanny valley" effects.
This also allows you to cross check the input name against names used in pranks (e.g., Yuri Nator; Seymore Butts; Anita Tums; Anita Dikkenmee; Holden Mycock). You could then decide if a prank name is detected to add a little laugh after the name is spoken, for additional personalization.
That's true. Our pitch is more or less revolves around the first part of the audio. So, it is definitely possible to add Hi and the name instead of editing the message in between which makes the uncanny valley part more pronounced.
Thanks a lot for the suggestion.
We will implement this.
On one hand, this is a technological marvel and props to the person who made it.
That being said, the name sounds extremely artificially inserted, placing the video in the uncanny valley. At first I thought it was just because you were using non-English names which the model wasn't trained on sufficiently, but then it was even awkward saying "James".
When I see my name clearly copy pasted into an email trying to be fake personalized, it makes me hate the sales email. Whenever I see those emails I just read it as "I don't actually care what your name is, but I think you're stupid, so I can probably trick you into thinking this is personalized".
Thanks for the appreciation and the criticism.
I think, at least for audio calls this is very ear-catchy.
Instead of cold automated calls, if your name is personalised and the whole pitch is personalised according to the category (say 5-6 categories of people in a sales pitch), it will look good.
Currently, for the free-demo, we chose to use a voice that sounds almost similar to the original voice, not the actual module that the customers will get once they sign in which is much more advanced. But I understand that it was a mistake. We will add the actual module as the demo next time we show to the Hackernews community.
5 comments
[ 0.23 ms ] story [ 25.9 ms ] threadThanks a lot!
This also allows you to cross check the input name against names used in pranks (e.g., Yuri Nator; Seymore Butts; Anita Tums; Anita Dikkenmee; Holden Mycock). You could then decide if a prank name is detected to add a little laugh after the name is spoken, for additional personalization.
Thanks a lot for the suggestion. We will implement this.
That being said, the name sounds extremely artificially inserted, placing the video in the uncanny valley. At first I thought it was just because you were using non-English names which the model wasn't trained on sufficiently, but then it was even awkward saying "James".
When I see my name clearly copy pasted into an email trying to be fake personalized, it makes me hate the sales email. Whenever I see those emails I just read it as "I don't actually care what your name is, but I think you're stupid, so I can probably trick you into thinking this is personalized".
Thanks a lot for your inputs!!