Building Better Harems


The following milestones should be understood as broad descriptors of our main goals, and not as any sort of detailed breakdown of each and every minor step required to achieve them. The goal of this list is to show our community the direction we intend to take Harem in, and provide a rough sequential order of when the most desired features will launch.

Racially Diverse Models

Status: Currently in development. MVP successfully built. Estimated 60% completed.

Summary: Currently due to the lack of availability of racially diverse datasets, there is a stark difference in quality between our white and non-white models. We know our users want women of all races in their harems, and are working to overcome this by placing special attention towards curating and analyzing datasets focused on non-white models.

Measurement of Success: When our average non-white model output is equally photorealistic as our average white model output.

Full Body Models

Status: Currently in development. MVP successfully built.

Estimated 40% completed.

Summary: Currently we are able to generate nearly photorealistic full body renderings of our models both clothed and unclothed. However, they are not quite as high quality or photorealistic as our torso/head/shoulders model renderings. As our goal is to provide users a complete Harem experience in which any of their models can be inserted into videos from head to toe, and eventually even have videos of them created from scratch, we understand the importance of developing high quality full body renderings of our models. We are working on bringing our full body renderings up to the same high quality levels that our torso renderings currently have.

Measurement of Success: When our average full body rendering is of equally high quality and photorealism compared to our average torso rendering.

Deepfake Capable Facesets

Status: Currently in research phase.

Summary: This is the big Magilla, the feature everyone is waiting for. "When will I be able to put the girls in my Harem into a video?" In order to provide this ability for you, we need to be able to generate hundreds or even thousands of pictures of the same girl, all from different angles and with different lighting. This is a surprisingly tough challenge, but we understand it is the number one priority for our users, and it is our number one priority as developers. Once we have developed this we will begin issuing NFT "facesets" which can be used in conjunction with common Deepfake technology in order to insert one or more girls from your Harem into any video pornographic or otherwise that you would like to see them in.

Measurement of Success: When our algorithm is capable of generating a varied enough faceset of each model it produces that the faceset can be used to create a photorealistic Deepfake using commonly available Deepfake technologies.

Deepfake Capable "Bodysets"

Status: Currently in research phase.

Summary: Currently Deepfake technology that is available to consumers is largely limited to "facial overlays". In other words, you can Deepfake someone's face onto a video, but not any other part of their bodies. This makes sense as even for pornstars, finding hundreds or thousands of high quality photos of a person's naked body, all within the same rough time frame, and all from different angles with different lighting, is very difficult. However, this restriction does not apply to our algorithm at all. That is why once we have successfully trained our model to produce facesets, we will move on to training it to produce bodysets as well.

Measurement of Success: When our algorithm is capable of generating at least one hundred different photographs of our full body model renderings, with these photographs having different angles, lighting, and positioning.

Posenet Integrated Full Body Deepfake

Status: Currently in research phase.

Summary: Existing Deepfake technology is largely limited to "facial overlay" as detailed in the above post on Bodysets. However, some work has been done in generating the basic technology needed to extend Deepfake's overlaying capabilities to full body rendering.​ Once Harem's algorithm is capable of reliably producing photorealistic Bodysets, we will integrate existing full body motion and pose transferal technologies and existing Deepfake technologies, in order to provide our users with the ability to insert their Harem into the desired videos, from head to toe.

Measurement of Success: When our algorithm is capable of overlaying a full body rendering of one or more of a Harem's models into a video, in full photorealistic quality.



Humans Need Not Apply