Tengrai version 0.8.1 Tengrai version 0.8.1

Tengr.ai 0.8.1 release notes

Model improvements

Since the release of 0.8.0, we doubled our machine-learning efforts to deliver 0.8.1 in record time to https://tengr.ai.
The new model presents significant improvements compared to the previous version:

  • Greater diversity among the images created. All four images should be fairly different.
  • The prompt following was improved, but we are still working on improving it even further, both in the Seer and Servitor modes.
  • The model’s performance has increased by 20% (more speed improvements are coming, and we hope to introduce new GPUs soon).
  • Better artistic images.
  • We’ll address certain issues in a future update, such as distortion in character faces seen at a distance, text rendering quality, and others. However, we believe the current enhancements represent a substantial improvement, and therefore, we are releasing it now.
Shattered reality

User Experience improvements

Since launching Tengr.ai version 0.8 two weeks ago, we have implemented significant enhancements to improve user experience, including:

  • We have added a file system with pagination that shows 20 files at once. Each file contains the prompt, settings, and four images with their upscale.
  • We have added the most popular file format, PNG, as a one-click download. JPG and WEBP can still be accessed in 2 clicks.
  • Aspect ratio selection was improved.
  • Energy-saving mode was introduced to emotional input.
  • We introduced a light theme to enhance visibility in various lighting conditions.
  • The settings menu was added to group the different settings in one place.

Note: Some of the changes mentioned above were released earlier as hotfixes.

Energy saving

Coming in the next release (0.8.2)

  • We will be dropping the thumbnails in image generation view. All four images will be shown in detail without needing to click them, so it’s easier to compare them.
  • Negative prompt input as a separate text area (setting will be available for those who prefer the current, more flexible option)
  • Default input method selection (seer or servitor for now, but more to come later)
  • We fixed over 100 bugs in 2 weeks, but there are more to fix.
Robot in a lake - landscape
  1. Hello, I just found tengr.ai and I love it. I am wondering when does my free 100 image generations reset?

  2. I reread your article on Hyperalign and I want applaud your effort in this direction. Other generative systems will simply shut down an attempt at generating or upscaling totally innocent images. Extremely frustrating for the well intentioned user. Your approach looks like a great step forward in responsible AI generation… Great… Thank you

Leave a Reply

Your email address will not be published. Required fields are marked *