Catching Heat [v0.5.0] [deepglugs]

male protagonist, monster girl, 3dcg, adventure, fantasy, drugs

1. Extract and run.

v0.5.0

Changes include:

Visual asset updates (using HD model)

Rework ceres_justusfans_event_2 scene

Rework elena_boyfriend_fight scene

Rework elena_ceres_argument scene

Rework steve hex scene

Use vp9 with constant quality mode for 3D image stills

Train on 550px images resized/cropped to 512px to avoid stretching (blurring)

Add voice for Tess (modified from Ceres voice powered by Endymion)

v0.4.0

Visual asset updates (using HD model)

4 new scenes

New AI voice for Ceres powered by @EndymionVA

20 second cutscene (Euryale’s dream)

Fix: Display wrong image in Ceres/Elena pool scene

v0.3.0

Visual asset updates (using HD model)

Voice asset updates (new AI voice framework/model)

New AI voice for Elena (powered by @EmpressBlueVA)

2 new scenes

Minor dialogue changes in first few scenes

Additional narration to help with flow in first scenes

Minor character changes (Elena now has “brown_eyes”)

Add credits screen to main menu

Replaced some poses in Tess’ classroom scene

Switch to webp image format with lossless compression

Statistics:

148 images

15+ PetaFlop-hours of training

36k training images

Visual asset updates

Voice asset updates

Change to text in intro scene

1 new scene (with 2 variations depending on previous choices)

v0.1.1

Visual asset updates

Voice asset updates

5 new scenes (from 0.1.0)

Halloween themes (Euryale wearing dia de los muertes mask and some AI jack-o-lanterns)

Minor text fixes

How our system works: poses, tags, etc defined

This post aims to help define and describe our process for generating game content.

pose: A human pose. These are usually generated using AI from models created in software such as Blender or other model/pose creating software.

tags: These are text attributes that describe the desired image. By changing the attributes, we can control many aspects of the generated image that would otherwise be very difficult to control.

In a nutshell: Our system works by taking a human pose, combining it with attribute tags and shoving it through a custom neural network that takes the two inputs and produces images, scenery, etc.

This project aims to push the limits of AI to generate compelling adult-themed games. The benefits from AI generated content include rapid regeneration of visuals and content tailored to specific interests or styles.

Under development is Catching Heat, the first in a trilogy of games we’d like to produce.  Catching Heat follows the story of a succubus who strives to seduce a group of individuals and use the “lust energy” from them for some noble purpose (she won’t tell us why!).  Her goals are thwarted, however, by a mysterious creature that was initially “helping” her with her plan.

The research and experiments to develop the initial models is a time-consuming process and uses intense compute resources (lots and lots of GPUs training for months at a time).  For this reason, we ask your help.  If you like our content, please consider supporting us in generating new unique content and pushing AI and gaming to the next-level.