Skip to content

Latest commit

 

History

History
118 lines (118 loc) · 5.19 KB

CHANGELOG.md

File metadata and controls

118 lines (118 loc) · 5.19 KB
Raw

Changelog

1.17.x

  • Requirements compilation

1.16.x

  • Added TTS AI
  • QOL fixes
  • Added new setting: use proxy or not
  • Merged NSFW Detector with Analyzer
  • NSFW Detector updated to new detection function
  • Added outputs gallery tab
  • Updating Video Analyzer functions
  • New toast outputs for all cleaning functions
  • Batch Upscaler

1.15.x

  • Updated CUDA to 12.1
  • Added restart MultiAI button in clearing tab
  • Full new NSFW Detector (including Video Analyzer, Image Analyzer)
  • Clearing tab moved to Settings tab
  • New CSS and JS styles
  • Model preloading setting has been removed, because the function is useless starting from 1.14.x
  • Better VRAM Control (fixing #6 issue)

1.14.x

  • Updated model loading
  • Fixed some bugs
  • Classes changed to .py files
  • Updated requirements versions
  • Fixed CLIP very slow generating
  • Updated GPU(CUDA) cache clearing
  • QOL fixes

1.13.x

  • Improved code readability
  • Changes recommended by Ruff have been made
  • The log system has been rewritten
  • Added tooltips instead of some logs
  • Fixed a bug that prevented the launch of the Video analyzer
  • Now None is selected in Upscaler according to the standard, which removed the error at startup
  • Release of AI Detector
  • Full new h5 model for detector, trained by me - https://github.com/ShamHyper/Classificator_HUMAN_VS_AI
  • QOL fixes

1.12.x

  • Added [BETA]AI Detector - detecting AI or Human created images. Very crude function, a lot of false definitions
  • Some fixes of AI Detector
  • Added threshold for AI Detector
  • New H5 model
  • Now all models downloading from our studio server
  • QOL updates

1.11.x

  • QOL updates
  • Libs versions update in requirements
  • Removed logs

1.10.x

  • Fixed live settings for #3 issue

1.9.x

  • Fixed all bugs of version 1.8
  • Updating for requirements.txt old-version libs
  • Fixed requirements.txt (gfpgan requirements compability)
  • Fixed bug with tmp.png in MultiAI/.. folder
  • Gradio update
  • Fixed single video Video Analyzer (idk why and how I broke this function...)
  • New color theme - with .css file for custom theme in ../settings
  • Clear log in clearing tab

1.8.x

  • Fixed all bugs of version 1.7
  • All defs renamed for simpler use
  • Progress textbox for Video Analyzer cleaner
  • Added Settings tab. Now you don't have to go into config.json and change something with your hands. This tab will do everything for you!
  • Logging for Debug mode (../logs)

1.7.x

  • Fixed all bugs of version 1.6
  • Work on optimization has been carried out
  • Now all outputs are in the same folders for ease of use
  • Fixed a bug with an uncleaned tmp_pngs folder that appeared if the batch process of Video Analyzer was interrupted
  • Added separate class for models
  • Added Clear all outputs tab
  • Code optimizations for config
  • New lines for debug (config.json)
  • Threshold slider for Skip drawings or anime checkbox

1.6.x

  • New AI - Video Analyzer. Analyze Video like in Image Analyzer, but analyze per-frame of video and give you avg. percentages specs. of them
  • Fixes for debug launch-timer
  • Removed IceCream logging for GitHub. Now this only dev-version function
  • Better model loading
  • Better console outputs
  • "preload_models" option in config.json
  • Added batch for Video Analyzer
  • Lazy mode batch Video Analyzer (skiping per x frames)
  • Threshold slider for Video Analyzer

1.5.x

  • New system of loading models (35s preloading, but faster working)
  • Code optimization
  • Fixed lying debug timer
  • Debug-timer optimizations

1.4.x

  • Deleted preload models because this function is corrupted and unuseful
  • Added checkbox in Prompt Generator - randomize temperature of prompt. Temperature controls the level of "creativity" of the AI on a scale from 0 to 1, the default value is 0.7. The lower, the more impersonal the text will be. The higher — the more unusual the text will be. I set the value from 0.4 to 0.9 when enabling the "randomizer". This will help to avoid duplicate prompts when generating with the same text.
  • Removed clearing function
  • Removed clear_need in config.json
  • Code optimizations (20% improve)
  • Loading timer for debug mode
  • Added emojis for buttons to better understand the interface
  • News about Python 3.11+, details in the README
  • Fixed console warnings that slowed down the promt generator
  • Added a new AI - prompt generator. With it, you can generate prompt, complementing your own (examples in the screenshots in the README)

1.3.x

  • Optimization of the detector functions has been carried out
  • Added the ability to choose whether to load AI models during program startup or during the execution of a specific AI. For more information, see config.json

1.2.x

  • Optimization of the image analysis function
  • Fixed several inconspicuous bugs
  • Added the ability to enable asynchronous model loading. Later it was cut out in 1.3+

1.1.x

  • Improved the output of prompts in Image Analyzer
  • Removed spam in the console associated with the previously added debugger for gradio
  • NSFW specifier renamed to Image Analyzer
  • Added CLIP interrogator to Image Analyzer to generate an image promo

1.0.x

  • Added NSFW specifier for evaluating images as a percentage (porn, hentai, drawings, netural, etc.)
  • Release. Beta changelog will be introduced later