I am building an SDK to support custom partner engine's to help devs further customize their partners and how they interface them into different stories, scenarios, worlds, etc.

From a technical perspective, this was a fun one to work on. I had to add a `partner.engine` attribute to support the dynamic creation/interactions with the partners.

For example, if `partner.engine` is anime-v0, it effects how the backstory, personality, image generation, etc runs.

Your voice is absurdly smooth and easy to listen too. Great video : )

Seems pretty interesting. I think if this is to become the standard of decoupled LLM tools/context, a lot of money could be made by simply putting existing APIs in the MCP format.

lol took the words right out of my mouth

Actually interesting. I think everyone assumed it was chat bots. I know I did!

Now when chatting with your partner, you can enable voice mode on and you get fully unique voices that sound.. in my opinion.. excellent! Sadly bluesky doesn't support audio upload at the moment, so go try it for yourself @ app.hoshi.chat

So the general flow works like this. On AI partner creation after we've successfully generated a fully unique profile with flushed out personalized/randomized characteristics, we allow an LLM to mix some premade voices together to randomize the voice based on the profile provided.

That is where www.cartesia.ai stepped in. As far as I can tell, there is really no limit to the amount of voices you can have & they offer really great voice editing tools such as voice mixing using multiple sample voices. Also, the prices are just absurdly better.

@elevenlabsio.bsky.social has the best voices out there. Hands down. However, the pricing model and limitations on the number of custom voices kills this idea in its tracks. How can every AI partner have a totally unique voice if, for $300 a month, I can only have 600 voices in my account?

Woah. That is awesome!

It really is the next paradigm of progress for LLMs