Littel HogeAI update. I was extremly hard working this week and got a lot of things done - here is a quick summary.

11 Jun 2023, 22:40
Littel HogeAI update. I was extremly hard working this week and got a lot of things done - here is a quick summary new bot implementations: - watermark function - overlay a whatermark file with transparencys - have the overlay scale and adjust automaticcaly to different image sizes - external settings file - Bot can load settings with image generation related parameters like schedulers, cfg, sizes, steps ... and many more - admin user implementation - admin commands - advanced text handler update - added override settings in prompt function - user can add parameter to prompts like seedinfo- prompt and rawpromt and text striping improved - - image captions introduced (text attached to the generated image) - including the seed parameters - if those are copied and put in as a new prompt (with the seed) an nearly identical image gets reproduced - general improvements and bugfixing - shared dictonaries, multiprocessing unit, response handling - tested out a nsfw filter got it installed but it was censoring to much - i have to find a bettter one that is adjustable - wildcards updated with HogeAI character prompts content creation: - develloped a Hoge mascot character - created various showcase contend - - started learning and did some first experiments with video and ai - did experiment with 3d depth channel extraction to calculate 3d stereoscopic image pairs model: - last trained (HOGExArtistmodel) model was tested and refined to a stage where its usable but could need more improvement - HOGExIllu model still makes the nicest outputs for my taste - but has its own feel - HOGExDeli model model lower res but quite flexible has more wider range usecases but image quality is less fidelity then the newer trainings due to beeing stabel diffusion 1.5 based so this is a quick summary .. might have forgotten some things. Next week I want to continue addressing the nsfw thing and do more bugfixing - also read more into encryption and save data exchange. We defentyl need more testing with the new implementations so this will be part of next tessessions. Also wanna concentrate on creating animation and play around with different techniques and forms of compositing.