I have been busy building a website called, sayit.love it is using OpenAI’s realtime voice API to make function calls to control your device (currently only works with Intiface) while it’s also giving you compliments.
Backstory:
I have watched a YouTube video that Cloudflare has published a while ago. He was using OpenAI’s function calls to control a robotic hand via Bluetooth.
I though that was super cool, and wanted to try myself, but I don’t have a robotic hand just laying around my house, but I do have a Kiiroo Keon, so… Here we are.
If you were wondering what was in the video, here is the link: https://www.youtube.com/watch?v=TcOytsfva0o
What I have done so far:
Integrated OpenAI realtime API (current version: gpt-4o-mini-realtime-preview-2024-12-17)
Added some random nsfw videos saved from civit.ai that will play in loop while your device is in use… (i will eventually replace them with videos i have received permission for)
Implemented a credit system and integrated it with Stripe. OpenAI is charging me per token, so I needed to make something similar. The pricing is not yet finalised, I do want to make a bit of profit, but not too much to piss people off. (will keep testing, and if anyone buys credit will be considered as a donation until the site is open for the public, price might be subject to change.)
I have generated some scripts, you could call them funscripts (but not really), categorised them in a database, so AI will change them based on what you ask for (faster, slower etc..).
Current challenges/limitations:
First I need to make sure I am fully compliant with OpenAI’s terms of service. If i can’t be I need to look for alternatives, currently it is not allowed to try to make a sexual roleplay with openai or anything like that, openai api was told it is controlling a massaging device and it should make compliments, like how you deserve this time for self care etc.
I’m not planning to launch an app for phones. This will stay web based.
Currently you are responsible to run your own Intiface server, it should work with devices Intiface works with, but there is no guarantee.
The videos im showing in the loop are not that exiting. I probably should load videos from pornhub or something…
Whats next:
Honestly I dont know, it’s up to you guys, what feedback I’m receiving.
I might just close this project down, i have it for my own personal use, i dont know.
But what I was thinking:
Offer all the voices OpenAI offers,
Ability to customise AI…
Anyways, i have been writing this introduction of myself and my project without the use of AI. its honest, and its full of errors. Register for my waitlist and try it out!
Creating an account only to advertise feels disingenuous, as does a lot of the service being vibe-coded.
If you’ve been part of the community before launching this service it would be more transparent to advertise using an active account with history.
If you haven’t used the forum before, it raises a lot of skepticism about how and why you sought it out.
Given recent activity, it’d be easy to assume this is yet another low effort AI cashgrab trying to dishonestly prey on forum users. Perhaps by the same person circumventing the ban.
Bottom line; there’s too much unpolished LLM output and your approach to advertising comes off as predatory.
Please make a logged-out test site that allows you to test it with a predefined set of cached responses
I.e. like a visual novel rather then an AI chat
(and honestly I’d prefer a carefully-crafted VN to an AI chat any day… maybe beccause I’m not really into AI chats yet)
AI products are really questionable atm so you have to prove well it works before we allow to keep this thread on site (we did delete a bunch of random AI-generated trash already)
will do thanks, in the meantime i want to emphasise that this project is kind of dumb at the moment, like our lovely friend @Rose said, but it is to be improved with your help.
So… I tried it out. I have to say. I’m not only NOT impressed, I’m rather disappointed.
I think the biggest thing setting you back is the fact that you’re complying and using OpenAI to run this. OpenAi’s AWFUL GPT4 makes everything sounding remotely nice sound like siri trying to talk dirty. It’s so incredibly robotic, stilted, and unappealing. (especially when they read out the whole script name and numbers included).
i found my old account so i’ll be posting from this one from now on, since someone called me out for using a brand new one. @Rose
first off, i wanna address @Toothlessninja’s concerns:
this feature is still experimental. since i added custom prompt support, it’s possible to get the openai model to talk dirty, sound better, even moan. but i think you were using the mini model, which is dumb af. it works sometimes, but most of the time it kinda sucks. the main issue is the cost of the non-mini version. i’ve made custom prompts that i actually enjoy for personal use, but yeah, it’s not cheap. honestly, it’d probably be cheaper and better to just send a control link to a cam girl in a country where this kinda thing is affordable. just being real.
in the meantime, i’m working on adding cheaper alternatives. they won’t be real-time, there’ll be some latency, but at least they’ll be more affordable.
i’ve also added some free features (which still need improvement), and i plan to add more for users who don’t wanna spend money.
at the end of the day, i made sayit.love to create something valuable. i really hope it works out and people enjoy it, but worst case, i’ll just keep it running for myself lol.
You can now stream your webcam, it detects your face, the more you open the mouth the more it will vibrate or stroke.
All processing happens locally in your browser, no image is being sent to us.
While it was designed to be used with your own face, nothing is stopping you to use software like OBS, to create a virtual webcam and stream your screen instead, like I did it here:
I can picture a scenario where some users open their mouth due to stimulation, like to moan, and get caught in a loop of ramping up intensity and mouth expansion xD
I tried this but the free credits arent even enough to get the device moving, I was able to get 2 voice lines out of the AI, and then it went to negative and told me i had to pay rip
yeah, i agree that feature is not cheap! hopefully ill be able to cut down on the cost. also planning to introduce other AI models, that might not be as “real time” as this one, but it will be much cheaper. in the meantime im also adding free features that do not require any credits, because those features dont cost me anything either, but they’re still fun.
