Skip to main content

I wanted to know. Does the AI chatbot in videoask work like chatgpt? as in if the responder asks anything from it, then would the avatar reply just like chatgpt? or do we have to put in the keywords ourselves and store in that information then it performs like that?

Hey @rameez noor, good question. The AI chatbot allows you to set certain keywords prior in the setup that are pre defined. Then what video segment (or URL) it re-route to if that keyword/s is picked up in the response. This will allow the respondent to describe verbally their answer and it will pick up on any keyword/s that have been set previously. 

I find it works well if you wanted to use it for FAQs. In theory if you got creative and had lots of different keyword options set, such as ‘pricing’, ‘how it works’ etc. Then you could set it to be very  intuitive.

Otherwise you could have the person on the video prompt very specific options to choose from. E.g. a question could be ‘please tell me your favourite colour’. This of course could be set up as multiple choice instead, but let’s say if you had so many colours that you didn’t want to display on the screen. This would allow it to feel more conversational and intuitive to the respondent.

You can set a fallback pre recorded video segment if it wasn’t able to pick up certain keyword/s in the response. This could say something like… “Sorry, I didn’t quite get your response there. Please try again or check out the FAQs page on our site”

 


@rameez noor I will add that I have been doing quite a lot with linking VideoAsk to different AI models through automation tools. Essentially if you are receiving data from your respondents, then you can pass that information received from the VideoAsk to do some pretty wildly creative and impressive stuff. I use Make.com predominantly to make the further journey very very customisable for business owners. The last option I did was for a legal firm that wanted to automate the generation of conveyancing quotes. This didn’t have an AI component to it, but what I am building at the moment does. 
 

The biggest limitation for getting an AI twin / personal avatar at the moment to draft a response back that has been dynamically generated from your previous responses is the time it takes to generate (and confirming it is possible yet through an API). I am definitely exploring this space to see what is possible and can keep you updated, if that is what you are wondering about.
 

In theory, you might at best be able to generate a personal avatar / AI digital twin that is created in response to your questions, then embedded in an email perhaps. This would give some more allowance for the time it takes to generate one. I haven’t got a proof of concept on that one yet, but the tech definitely now exists.

I hope that helps!


@Ash.NZ So when it picks up the topic or the keyword we have put in or anything else, do we have to put in the replies for keyword ourselves or will ai do that itself and generate a reply itself just like chatgpt? I think the only thing their AI does is that it picks up the keyword itself right? it does not work like other chatbots that we see around right?


@rameez noor correct in your final statement. To clarify, it will only pick up on the keyword to redirect, not to generate a brand new video response that directly relates to their inputs.

I am doing some work at the moment to generate a text summary that can be featured alongside the VideoAsk. This is generated with the assistance of AI as a completely separate custom solution that has received responses and interpreted data (in essence, a chatbot that has been helped with the assistance of VideoAsk at the start). The idea being that the video can say something like here are some of our recommendations for you (I have a proof of concept for this one). But unfortunately the tech at present is limited by how long AI videos take to produce and add. That might change over time. Maybe if you had quite a long VideoAsk that extends beyond the time it takes to generate a a new video, it may be possible to insert a summarised AI driven video response / summary at the end that has interpreted the user’s inputs at the start. But that would be the best outcome I can conceptualise at present. And it would require setting up a very specific make.com scenario that links all the data and technology together. 

 

Hope that helps!

 

 


@Ash.NZ Thanks alot Ash for the great replied. When do you think these videask avatar videos will work like an actual chatbot? are they working on it or something? will it come anytime soon?


No problem. The biggest issue is the time it takes to generate. I am sure this will improve over time and eventually this will become “a thing”. I would imagine by that stage it will be adopted en masse. As for when that would be, I have no idea. 

At the moment I think the best we can do is generate a newly generated and tailored response at the very end to follow in an email. So this would be the best option we have at the moment for a workaround. It would require using an API for the likes of Synthesia and using an AI module to interpret the user’s answers/intent and then prepare a script for the avatars response to then be generated from. 


Reply