@Valentin Ffrench Hmm, now that I started researching, the ElevenLabs does not offer a direct outbound calling functionality. It seems Twilio or similar 3rd party services should be integrated to place the outbound calls using the audio generated by ElevenLabs. Which means additional cost per call, which varies across different regions (https://www.twilio.com/en-us/voice/pricing/us - as example, some Asian countries are up to x20 more expensive than USA). In such case, it would be difficult to estimate the operational expenses for 3rd party/Twilio API usage per month, which are incurred on top of ElevenLabs charges. Overall, this is not ideal, which forces to reconsider the first approach - capturing customer details during a call, and having customer initiate the call via ElevenLabs widget integrated into the website

@Valentin Ffrench Thank you for all the suggestions, Valentin! I'll give it a try - the ElevenLabs x Twilio. Actually I recalculated the costs, and they seem to be ok - for 1,100 minutes per month (18 hrs of outbound calls, or 110 calls of 10 mins each) we will spend $99 for ElevenLabs and approx. $44 for Twilio - overall it still makes sense, even if many calls will not convert and will be inquiries only. Thank you for sharing the video from Bart Slodyczka. In terms of Twilio cost for outbound calls, I averaged the outbound call cost over the selection of out top client countries

Hi @Nate Herk this is awesome. I was wondering if you'd be interested in experimenting with a more complex 'human-in-the-loop' use case involving AI agents—specifically for serial TV writing and screenwriting. This would be one of the most context-heavy applications out there. As a screenwriter, I’ve often felt that the creative process could benefit from integrating AI agents. What I’m thinking of testing is a setup where five separate AI agents are each assigned to one of the five main characters in my series. Each agent would be given their character’s backstory and personality traits, and we'd use an n8n workflow to collaboratively brainstorm the most emotionally charged, coherent plotlines. - Plausibility: Characters behave based on their evolving emotional histories. - Efficiency: An overlord agent filters out incoherent or irrelevant ideas before human review. - Writer Control: Humans can override AI suggestions when logic or emotional tone falters. Would this be something you’d be interested in exploring—as both a test and a learning opportunity—using a real-world TV series setup? If this works, I would be glad having a human cowriter :)

@Nate Herk Fought with it for 2 hours and can't seem to be able to setup Tavily HTTP request :/ method: POST URL: https://api.tavily.com/search Authentication: None Send Headers enabled Specify by JSON [ { "name": "Content-Type", "value": "application/json" }, { "name": "Authorization", "value": "here is tavily API key" } ] Send Body enabled Specify Body using fields below Name -- if I add static JSON expression it formats it in a wrong syntax so that Tavily doesn't get a single string expression hence won't run a search. workaround advised by chatgpt is too complex Value -- by model?? If I let model specify the entire body there is a tool schema mismatch error much more complicated than I thought there is Tavily in custom n8n nodes coded by someone (link video) but I would rather figure out how to setup the POST request

@Nate Herk Thanks Nate, actually the easiest way was to open Jarvis template and check how you set it up (it works now). It was a good experience to try to figure it myself though ))

Hi everyone, Andrew here. I'm a hospitality professional based in Sri Lanka learning AI, architecture, and doing screenwriting in my spare time. I am grateful Nate established this community. I am looking to build a personal assistant which will know my personality to aid in numerous design and screenwriting projects. I am also curious to build & apply AI agents for some of the travel agency company workflows.