Sign In

TMNT Turtles AIO

93

611

10.2k

17

Verified:

SafeTensor

Type

LoRA

Stats

611

10.2k

71

Reviews

Published

Nov 23, 2023

Base Model

SD 1.5

Training

Steps: 2,800
Epochs: 10

Usage Tips

Clip Skip: 1
Strength: 1

Trigger Words

TMNTGroup
TMNTDonatello
TMNTLeonardo
TMNTMichelangelo
TMNTRaphael
a group of teenage mutant ninja turtles
a teenage mutant ninja turtle
wearing purple bandana mask with eye-holes
wearing orange bandana mask with eye-holes
wearing red bandana mask with eye-holes
+7 more

Hash

AutoV2
9ACC7A53D1

This is a top down rework of my previous TMNT Turtle models, now rolled all into one convenient LoRA. It heavily favors photorealism but seems to work decently for cartoon/comic book styles too. This model was trained on a combination of about 300 images including different art pieces/styles and select movie frames from the first 90s TMNT movie. I haven't thoroughly tested it but I can give you a break down on how it was trained, the keywords I recommend, and my experience with its performance so far. It struggles with groups, can do duos, and Leo consistently requires the mask related keywords discussed below. - (Note: I have not successfully been able to generate more than 2 turtles at a time and even if I do they are identical, despite the training data having ample examples and tags of all 4 brothers together in many scenes)

Recommended keywords/tags:

In the event that the Turtle you are prompting for does not generate properly- i.e the wrong mask color- I highly recommend using the phrase "wearing (purple/orange/blue/red) bandana mask with eye-holes" but without the ( ) and /

Key triggers are going to be:

  • a teenage mutant ninja turtle

  • nunchucks

  • wooden fighting staff weapon

  • sai dagger weapons

  • sword, swords, sheathed swords on back

  • wearing ________ bandana mask with eye-holes

  • TMNTDonatello

  • TMNTLeonardo

  • TMNTMichelangelo

  • TMNTRaphael

  • TMNTGroup

Training:

About 300 images of different sizes trained at 768 res

Repeats: 3

Batch size: 3

Epochs: 10

Text encoder rate: 0.0001

Dim: 64

Alpha: 32

Model: base SD 1.5

Everything else was standard settings from the holostrawberry LoRA trainer colab.

If you enjoy my work, please Like/Review/Upload your generations and if you're especially interested in supporting my work you can tip me some Buzz or support me on DeviantArt/Patreon/Discord, I accept commissions for both Ai Assisted art and LoRA model creation as well as donations to assist furthering my projects. I don't have substantial cash flow to support pumping out models and am on an AMD Win 11 PC so training isn't an option locally, so if you're interested in assisting me further or want something special, let me know.

DeviantArt: https://www.deviantart.com/archangelaries

Discord Channel (help me build a community and friends): https://discord.gg/Cj42kGze

Patreon: patreon.com/ArchAngelAries