Hi Jannis, you’re my favourite creator in the voice Ai space and for me, as a little bit younger guy, also a huge inspiration! It would be incredible valuable if you could launch a full guide on how to build an appointment setter, that is able to do all these things (Bookings, Cancellations, Resceduling)! I’ve seen many tutorials so far and I think you’re one of the most skilled and knowledgeable guys in the whole voice ai space I’ve seen so far. Also your old Restaurant Table reservation assistant tutorial was very good, but sadly it didn’t include the google calendar stuff and its also a little bit older. I think the whole voice AI space has huge potential in the future, I am from Germany and here the voices aren’t completely realistic. I know a video like that would take a lot of time, but even if you’d charge for something like that, I think many people would purchase. Thanks a lot for your work and good luck in the future!
Great overview of the huge boost that Vapi and other providers are getting with these updates. Thanks for the insightful share and your perspective on this opportunity.
Amazing explanation Jannis! I have way too many projects on my plate right now but watching this is making me want to add vapi integrations too haha...
Sam Altman told everyone, do not build companies around the features of GPT or you will be made obsolete, literally said it… You want to use the power of GPT to actually offer something novel
True that. Altman advised companies to innovate beyond basic integrations and focus on building unique, defensible products with long-term value rather than simply relying on the immediate functionality of existing models.
Amazing video! With all your knowledge of using VAPI and then also leaving VAPI in tools call using make to actually handle bookings or other requests through webhooks/API's - do you think VAPI will integrate a layer to tool calls with easy integrations to post or receive data to e.g databases. Or will we still need to use third party SaaS to lookup eg. customer information in databases and send information back to the VoiceAssistant ? I would love a video on the discussion of that subject, because one thing is the capability of the technology the second part is the usability for our customers. Thank you so much for all the other videos and for being a pioneer in communicating the voice development as it happens day by day.
Your video is very cool and informative. The only thing I cannot confirm is the thing with empathy. I’m working since over nine years in psychology and we are about to give our best to implicate this empathy and we are 100% sure that AI has the ability to trigger and create emotions. ❤
Yep, empathy is still wacky, but it's getting a lot better. The fact that it can actually tell Jokes without simply talking monotone is already quite impressive. I believe it's not too long anymore until we have our first empathetic conversations. :) Very exciting times!
Excellent explanation! Really informative. I would like to know if I can deploy my own services using the schema shown at 1:27. I want to deploy my own solution, but I would appreciate it if there is any material available, preferably the solution I want to deploy should work good in Spanish. I really appreciate any help or guidance you can provide.
Hume AI has had voice to voice demo available for a while yet it actually seems to fail on some of the issues you’re proposing it will fix, eg returns inappropriate tone, interrupts you more often assuming you’ve finished speaking, sped not noticeably different. Agree on its potential but yet to see results.
Hume isn't a native speech to speech from what I've seen so far. It still goes the same orchestration route, just with emotional recognition. I like their approach, but it feels still very clunky
I don't like Vapi and realtime api pricing is abusive. Can you show us how to do: Twilio input > Webhook > Deepgram STT > Memory + AI > Deepgram TTS > Twilio output
Why would you build with yesterday's stack? I agree about pricing, but Google and others will be snapping at their heels (as evidenced by NotebookLM). I doubt they can keep up their current pricing for long. My guess is that it is more of a provisioning thing. They don't want (are not ready for) mass adoption just yet. I still plan on building with it, but realize it will not be so attractive till the price drops. If you really want to to use yesterday's stack, have you checked out Groq? Their super fast inference makes up for a lot.
but if you want it in production you need to have a dedicated server that uses websocket clients to connect with the deepgram websocket etc... that's when vapi comes in
Vocode might be the right solution for you then. I don't recommend starting your own structure though, except if you have the resources to throw a dedicated team on it.
Hi Jannis, you’re my favourite creator in the voice Ai space and for me, as a little bit younger guy, also a huge inspiration! It would be incredible valuable if you could launch a full guide on how to build an appointment setter, that is able to do all these things (Bookings, Cancellations, Resceduling)! I’ve seen many tutorials so far and I think you’re one of the most skilled and knowledgeable guys in the whole voice ai space I’ve seen so far. Also your old Restaurant Table reservation assistant tutorial was very good, but sadly it didn’t include the google calendar stuff and its also a little bit older. I think the whole voice AI space has huge potential in the future, I am from Germany and here the voices aren’t completely realistic. I know a video like that would take a lot of time, but even if you’d charge for something like that, I think many people would purchase. Thanks a lot for your work and good luck in the future!
Appreciate the input, and glad you find my videos helpful!
I’m working on something big that will definitely help with that. More news soon :)
Amazing like always. Perfect Deep Dive!
Fantastic explanation, great value. Thank you!
Great overview of the huge boost that Vapi and other providers are getting with these updates. Thanks for the insightful share and your perspective on this opportunity.
Thanks, Trejon! Appreciate the feedback.
Amazing explanation Jannis! I have way too many projects on my plate right now but watching this is making me want to add vapi integrations too haha...
Banger after banger Jannis!
Thanks, Brock! Looking forward for your first custom-coded Realtime API app :)
Great explanation! Do you provide the Miro board links to these diagrams as Zooming in on the images your Hub gets blurry and cannot read.
Thanks for sharing your vast knowledge Jannis.
Sam Altman told everyone, do not build companies around the features of GPT or you will be made obsolete, literally said it… You want to use the power of GPT to actually offer something novel
True that. Altman advised companies to innovate beyond basic integrations and focus on building unique, defensible products with long-term value rather than simply relying on the immediate functionality of existing models.
Looks like facebook era. Also A way to discourage competitors imo. Do average pro know de how to deploy any gpt app ?
Jannis is the ONLY way to go when it comes to Voice Callers and Automation
My man never misses how are you pumping so hard man
I’m surprised myself
Great video!
Alll over it. Great video
Thanks for the video. It makes a lot of sense.
Amazing video! With all your knowledge of using VAPI and then also leaving VAPI in tools call using make to actually handle bookings or other requests through webhooks/API's - do you think VAPI will integrate a layer to tool calls with easy integrations to post or receive data to e.g databases. Or will we still need to use third party SaaS to lookup eg. customer information in databases and send information back to the VoiceAssistant ? I would love a video on the discussion of that subject, because one thing is the capability of the technology the second part is the usability for our customers. Thank you so much for all the other videos and for being a pioneer in communicating the voice development as it happens day by day.
Your video is very cool and informative.
The only thing I cannot confirm is the thing with empathy. I’m working since over nine years in psychology and we are about to give our best to implicate this empathy and we are 100% sure that AI has the ability to trigger and create emotions. ❤
Yep, empathy is still wacky, but it's getting a lot better. The fact that it can actually tell Jokes without simply talking monotone is already quite impressive.
I believe it's not too long anymore until we have our first empathetic conversations. :)
Very exciting times!
Great video! (Cooking up something similar and this helped a lot with research)
Thanks, Hugo! Keep up your great content too!
Excellent explanation!
Really informative.
I would like to know if I can deploy my own services using the schema shown at 1:27. I want to deploy my own solution, but I would appreciate it if there is any material available, preferably the solution I want to deploy should work good in Spanish.
I really appreciate any help or guidance you can provide.
Hume AI has had voice to voice demo available for a while yet it actually seems to fail on some of the issues you’re proposing it will fix, eg returns inappropriate tone, interrupts you more often assuming you’ve finished speaking, sped not noticeably different. Agree on its potential but yet to see results.
Hume isn't a native speech to speech from what I've seen so far. It still goes the same orchestration route, just with emotional recognition.
I like their approach, but it feels still very clunky
Is it cheaper on Vapi compared to open ai.. why Vapi is better then directly using it from OpenAI
It's not really about the price, but the utility - that's precisely what I explain in this video
@@jannismoore exactly ! vapi next version with realtime API will be a killer product !!
It’s $15+ hour for realtime API output.
Too expensive to use in a product except for wealthy clients.
Nice video, thank you.
You'll see those prices drop faster than the WiFi signal when you step one foot outside your house
@@jannismoore 🤣🤣
I don't like Vapi and realtime api pricing is abusive. Can you show us how to do:
Twilio input > Webhook > Deepgram STT > Memory + AI > Deepgram TTS > Twilio output
Why would you build with yesterday's stack? I agree about pricing, but Google and others will be snapping at their heels (as evidenced by NotebookLM). I doubt they can keep up their current pricing for long. My guess is that it is more of a provisioning thing. They don't want (are not ready for) mass adoption just yet. I still plan on building with it, but realize it will not be so attractive till the price drops. If you really want to to use yesterday's stack, have you checked out Groq? Their super fast inference makes up for a lot.
but if you want it in production you need to have a dedicated server that uses websocket clients to connect with the deepgram websocket etc... that's when vapi comes in
Vocode might be the right solution for you then. I don't recommend starting your own structure though, except if you have the resources to throw a dedicated team on it.
Many people underestimate specifically that part A LOT
Prices will definitely drop, and I don't think it'll take long either
Now thats an educational master class On the Voice Space. Thank You Jannis 🫡💪🎯🎬🦅
Must be getting paid by Vapi
Did it ever come to mind that we just love using it?