Home Artists Posts Import Register

Content

This is a small update mostly for the AI camera feature in Alive but it includes also a few small fixes and other new things.

AI CAM


This update includes some new features, settings & fixes for the AI Cam app. For people out of the loop about what's that about, it's an offline feature that allows connecting VAM to Automatic1111 (UI for StableDiffusion) that must be running in the background. When connected it allows processing VAM screenshots with AI directly in Alive, making them much more realistic. Besides screenshots, it can be used as a general-purpose tool to control any kind of AI images by allowing setting up scenes in 3D space or VR through VAM.


CONTROLNET

ControlNet is a local feature but it requires a bit of setup for people that want to use it. Long story short, you have to install an extension in A1111 then download some controlnet models. A quick ELI5 guide on reddit can be found here, also many videos on youtube . Before trying it in Alive, you should make sure you that you have it installed in A1111 and it works there. When ControlNet is activated (from Prompt > Photo type) generating images will be considerably slower.

ControlNet (creative) is txt2img + vam screenshot as controlnet reference. In this mode the AI won't care at all about the colors from VAM, just the shapes and your text prompt.
ControlNet (precise) is img2img + vam screenshot as controlnet reference. In this mode the AI will also try to follow the colors from VAM. Using the image creativity slider will let the AI be a bit more creative with the colors. This mode is similar to the Normal mode, but it forces the poses and shapes to stay closer to what the AI received from VAM.
Normal - is img2img. This is the fastest process but also the most loose and the hardest to control out of the 3. It's ok for enhancing screenshots, but not great for building AI images with more variety that you can control from text, like changing backgrounds,clothes, etc. You're kind of stuck with what the AI sees from VAM. Increasing the creativity will make the AI 'hallucinate' in random directions and you might lose details that you wanted to stay in the shot, like the model pose. You can see an example on the differences in this video.


DESKTOP SCREENSHOTS ROTATION

This is a feature I always thought was missing from VAM: the ability to rotate the photos angle on desktop. Initially I added it for the AI Cam but I ended up adding it to VAM's default screenshots as well. To rotate the screenshot camera you can press Q and E. 


VR UI

I changed the Aeternum UI button in VR to be a long-press, to avoid interference with other VAM features and scene buttons. To toggle it on and off you have to hold A or X for about 1 second. I was meaning to change the way the UI works in VR and try to mimic exactly how the vam UI works, dragging, floating and all that. But due to the button being already mapped to other stuff it's proven to be very tricky. I might try to do it as a combo of trigger+(A/X), or trigger+(B/Y) in the near future.



CHANGELOG

  • CHANGE: AI Cam changed screenshot logic to be more similar to vam's screenshot taking process; removed old preview logic that was bad for performance 
  • CHANGE: AI Cam added custom width & height logic 
  • CHANGE: AI Cam added ControlNet support and two controlnet modes (txt2img & img2img), image type settings
  • CHANGE: Added error message on A1111 errors and blank image result instead of remaining in infinite loading 
  • FIX: sanitized AI Cam prompts to prevent A1111 errors on some characters (like new line)
  • CHANGE: changed AeternumUI button (A/X) to toggle the UI on long-press (1 second), not on any press, to prevent interference with vam UI buttons and selecting things in scenes
  • NEW: added support to rotate VAM screenshots on desktop with Q and E, works with AI cam as well


INSTRUCTIONS

- Alive.cslist  - add to person
- Aeternum.cslist - add to session for the UI

The Alive Home scene needs Aeternum to be loaded to bring it to life (buttons, sky, cars, etc).

Alive doesn't need internet access, but it uses the web browser to render the UI in it so a few settings might be needed to get it to work. In User preferences Security these need to be enabled or there might be errors: Allow Plugin Network Access, Enable Web Browsers, Allow All Sites.

If you want to keep the sites whitelist enabled, you can add this to your whitelist.json file in vam's main folder, to whitelist just the UI file needed for Alive: "Saves/PluginData/Aeternum/Ae/Html/mainmenu.html"

AI STUFF

These are optional for people that want to add more life to VaM. By default Alive doesn't require any of these but they can be activated from the Services app.

  • Local text AI (Chat) - requires OobaBooga (quick guide)
  • Local image AI - requires Automatic1111 (quick guide)
  • Local Text to speech & speech recognition - require my Text Audio Tool (quick guide)
  • There's also support for online ElevenLabs AI voices, that requires an account on elevenlabs

CREDITS

Assets used: https://drive.google.com/file/d/1M_F6iWz1zK1Av2hgLNsR-6WOrBOqJs5s/view (3d objects from sketchfab, music from soundcloud)

Ashauryn - morphs ( from AshAuryn.Expressions.2, AshAuryn.Pose_Tools.2 and AshAuryn.Sexpressions.2, AshAuryn.AshAuryn's_Female_Genitalia_Collection.2, AshAuryn.AshAuryn's_Assorted_Expression_Pack_with_75_Morphs.7, AshAuryn.AshAuryn's_Tears_and_Pain_Pack.3)
klphgz - poses ( from Crouching, Standing, Lying & Prone, Sitting packs)
Jackaroo - morphs ( from Jackaroo.JarModularExpressions.2)
kemenate - morphs (kemenate.Morphs.10)
geesp0t - spankings vr hand collision detection (Spankings Plugin Mod v3.6 on reddit)

MacGruber, VR Adult Fun, Acid Bubbles & other plugin makers whose work inspired me to make some of the features here and plugins in general.

ElevenLabs.io AI voices used often in my videos.
OobaBooga's text-generation webui & AUTOMATIC1111's stable-diffusion-webui


FILES

SPQR.SPQRAlive.46.var 


Files

Comments

chris r

Any chance we'll be able to control eni noises like the cars....theyre just so loud to me :(

SPQRAeternum

yup, I have some stuff already but I wanted to release this faster for people that want to play with the AI images over the weekend. I'll likely release that next week

thomas d

I have not tried v46, yet, just wanted to comment my voices from the share site you mention in your earlier post expired, I could down others as well. I also noticed there are quite a few voices that can be downloaded from Windows 10, and probably other versions as well, go to Settings and Languages, Speech and add voices, it just downloads them from Microsoft. There are quite a few in English. I tried a Spanish, voice and she spoke English with an accent. Anyway, combined with being able to slow down or speed up the voice in your script, gives many more free options for voices. I still use the Rivescript a lot, and my script has become pretty comprehensive with all my updates. I have not tried the other free voices you suggested for OB. I just wanted to point this out if others wanted to download free voices right within Windows. Maybe you have already stated this, but I do not remember seeing it.

SPQRAeternum

Thanks! yeah, likely I didn't mention it, I didn't look much into other voices, I already had some voices myself. I thought MS would be doing subscription based voices on their Azure platform, and that big corps in general kind of moved away from offline TTS a while ago. But I'll check them out too