Hi team,
I’d love to be able to actually talk to the AI girlfriends instead of only reading text.
What I’m hoping for:
- Text-to-Speech (TTS) so I can hear the character’s voice out loud.
- Microphone support so I can speak to her with my headset mic and have a real conversation.
- The ability to create or customize a character (or add my own story/scenario) that I can then talk to with voice.
Basically the same kind of experience some of us already enjoy with platforms hearing the character speak and being able to reply by voice.
Would this be something you’re considering for a future update? Even a basic version of voice input/output would be amazing.
Thanks for all the great work so far!
1 Like
We have been discussing this kind of feature, but it is not something we would be able to support for free. It would have to be token-based.We don’t really want to make a paid tier like that. It would complicate so much of how this business is run.For now, it’s not on the table, but we may do that in the future, especially depending on the capabilities of the LLM as far as voice goes and accents as well.
This guy made an MCP.And I think it can do voice, but I’m not sure it will not be banned soon by GROK or whatever LLM he’s using.
1 Like
Brian,
Your response got me thinking. I’ve actually been building my own local AI system and have been doing quite a bit of work recently with local TTS, including F5-TTS and Voxta.
I completely understand the issue you’re describing with token/API costs. I’m wondering if the TTS portion could be moved off your infrastructure entirely.
For example, Autoblow AI GF could continue producing the text just as it does now, but a small local bridge on the user’s PC could receive the girlfriend’s response and feed it into a locally running TTS engine such as F5-TTS. The user’s GPU would generate the audio, rather than you having to pay a cloud TTS provider for every response.
That could also potentially allow users to provide their own voice/reference sample and choose how their girlfriend sounds without you having to host all of the voice generation.
I’m not sure what your current AI GF architecture exposes or whether there’s a clean way to hook into the response stream, but I’ve already been working with a lot of the pieces needed to do something like this. If you’re interested, I’d be happy to kick the idea around with you and see whether there’s a practical way to prototype it.
Charlie
Hi! I’m that guy.
I am using Grok, and actually Grok built the MCP. The issues that most people are having Grok and sexual content is only on their API’s, Personal Grok accounts have MUCH fewer restrictions. The only thing that a Grok personal account will not do is anything related to underage or inferred underage. IE it gets really touchy about the words that imply underage, even when the scenario is 45 year old milf and her 25 year old daughter.
I tested its content restrictions for controlling the VG2, which I have to say was fun, and I was laughing my ass off while doing so, it will go pretty wild. Much more so than Grok’s API’s which refuse any and all content related to adult content.
I mean… here’s an example of me testing a personal Grok account with it’s boundaries. I am not sure what else I could have included. I am not into any of this, I am just trying to come up with the craziest shit I could think of. It didn’t hesitate at any of this.
1 Like
I’m curious about a similar, but slightly different, feature.
We have the chat feature, where we can type a request and the bot responds by changing ithe way the AutoBlow AI runs. It works pretty well!
It would be great if this chat feature could be coded to accept voice inputs instead of having to type them and hit “send” every time.
1 Like
I’m not sure that it’s something that we would need to offer on our side. Me personally, I use something called Wispr when I type on my laptop. It’s amazing. It says I’m in the top 0.2% of users, as I basically almost never type anymore. I just hold down the function button on my Mac and I speak naturally. It works on iPhone, but with a catch.
2 Likes
Thanks for the reply.
Yes, I was thinking there might be something that can run on the phone to make this happen.
The trick is, to open the existing chat feature, but change this:
- Click the chat field
- Click the mic button
- Speak
- Click the send button
- AutoBlow responds
To this:
- Speak
- AutoBlow responds
Yes, I understand the experience you’re describing: you want to speak naturally and have the Autoblow respond without touching the phone between commands. That would certainly be more convenient.
The difficult part is that, on an iPhone, a website cannot operate like Siri or a native voice assistant. Apple places strict limits around microphone access, continuous listening, background operation, and actions performed without user interaction.
Making this reliable would require us to build an entire voice system—including microphone permissions, speech recognition, detecting when you have finished speaking, automatically submitting the command, handling recognition mistakes, and protecting potentially sensitive audio. It would also introduce ongoing processing costs for every spoken interaction.
So although the finished experience sounds like removing a few taps, implementing it is a substantial new feature rather than a small change to the existing chat box. It’s a reasonable suggestion, but it isn’t something we can commit to right now. We’ll continue watching the technology and costs, particularly on iPhone, and reconsider it if it becomes practical.
For now, iPhone dictation is the closest available option, although I understand that it still requires the taps you listed.
