Last Days to Subscribe Before AI Chat Pricing Increases
Added 2023-09-30 16:47:43 +0000 UTCIf you already have a current AI Chat subscription, your price will not increase. New users will need to subscribe at a higher price.
The chat servers are getting full, and people are sometimes having slow responses. We need additional income to expand the servers.
We are expanding to include an additional voice service, with a variety of slower but better sounding voices (but without voice cloning). This may allow some of you to avoid $22-$99 per month fees for ElevenLabs.com. This will also distribute the load from our current voice servers.
We are also looking into additional ways to expand the Chat servers with larger (smarter) language models. This will also distribute the load from our current AI language model servers.
Our current monthly server hosting and development costs are far higher than the AI Chat income, so we will be raising the price, as we continue to also add value.
After adding the new voice and language model services, if delays seem common in the current main servers (primary US and EU servers) we will also add an additional main US server.
If you already have a current AI Chat subscription, your price will not increase as long as you stay subscribed.
Comments
I'll look into it, thanks for the suggestion! We are generally working on a very large update for many languages, but it may take a couple of months to complete it. Korean should be available then.
vamX
2023-10-06 02:08:16 +0000 UTCPlease add Korean !!!
성운 이
2023-10-05 11:03:05 +0000 UTCBed Poses, Standing Poses, and Chair Poses are very different from each other. Internally it's impossible to anchor poses into the same spot, but there is an option you can choose, a checkbox in the bottom of the pose tab "Load Male Poses". If you don't load male poses, you won't be moved at all when new poses are loaded, and you can just move your body into position.
vamX
2023-10-05 03:06:26 +0000 UTCCan anyone explain how to anchor all poses to the same spot? I’m having an issue playing in VR and using embody, every load just moves around, and 9 times out of 10 when a new pose is loaded my characters body is turned like a pretzel and usually all the collisions have misaligned things so it’s not even that affective Turing the character around and twisting the wrists back to the right place
Chris
2023-10-05 00:01:28 +0000 UTCWe will be working over time adding many additional languages, including Italian. We have new voices coming soon which can speak additional languages, including Italian, but the AI itself doesn't know Italian, so we must set up translation layers on either side to translate Italian -> English -> AI -> English -> Italian. So the Italian won't be spoken perfectly if NSFW (for SFW the Chat AI can speak Italian natively). In the future there will also be better NSFW AIs that speak more languages, and when that is available the Italian output will also be more natural.
vamX
2023-10-02 03:02:03 +0000 UTChave you think to add italian language for ai?
Hyt450
2023-10-02 00:50:35 +0000 UTCThe way that we are handling things requires our own servers. For example we have a custom AI set up to determine what vamX actions should be run, plus the servers do all sorts of proprietary things in order to make everything work, for example elevenlabs returns only .mp3 files, but they must be converted to .wav to work with the ai plugin in vam currently, we do proprietary processing to make the responses short enough to not take forever with audio responses being generated by servers, we process the personalities into things that the AI chat LLMs can handle, we run a server for the web interface, etc., etc. (it takes 3 servers to complete a minimalistic configuration, and we use 8 different servers for our full solution). We don't just sort of link an LLM to Virt-a-Mate. Also, in my tests, things are slower if you try to run an LLM + Virt-a-Mate at the same time. So, this is on our general to do list, but we will need to think through how best to do this as it's not a simple matter. It may be that you would still need to be using some of our servers, but could provide an IP address for your own LLM on your computer, or that you could use an LLM from your computer, but it would be limited (not generate vamX actions or have other code that we use for optimization).
vamX
2023-10-01 04:00:35 +0000 UTCI have a 4090 that I bought for AI/ML purposes (and the fact that it makes VaM run really well is a huge bonus) and I have a local install of LLAMA2 that runs pretty flawlessly and speedily using oogabooga, and I've got my own connection to ElevenLabs that I use for creating race recaps for my iRacing League anyway. When can I expect VaMX to introduce the option to have the AI solution run entirely locally for free (aside from the substantial cost of purchasing the top-shelf hardware, of course), the same way SPQR did with Alive?
TomahawkJackson
2023-09-30 22:37:40 +0000 UTC
