mmtp16 - "Male Miles Tails Prower 16 Vectors" embedding
Training using the Textual Inversion tools provided by https://github.com/AUTOMATIC1111/stable-diffusion-webui using an RTX 3070 with 8GB VRAM.
Redistribution of these embeddings is encouraged.

Training parameters:
- name : mmtp16
- initialization text : *
- trained checkpoint: yiffy-e18.ckpt [50ad914b]
- number of vectors per token : 16
- learning rate : 0.005:5000, 0.004:8000, 0.0025:10000, 0.002:20000, 0.0015:30000, 0.001
- Trained as a subject (object) using a modified subject_filewords.txt template

WARNING: Long Filenames in image data may fail to process in Windows without enabling long ntfs paths.
E.g. "00116-0-by coolblue,penis,nude,balls,blush,erection,tongue,open mouth,cum,anus,cum inside,blue eyes,legwear,sweat,furniture,markings,orgasm,black nose,cum in ass,bed,yellow body,footwear,orange body,orange fur.png"
Image data is provided as a ZIP archive for this reason. If extraction fails, try a shorter path / folder name on your storage drive.

Trained for up to 100,000 steps total.
Embeddings are provided at 500 step intervals.
Training image data entirely sourced from e621 and tagged using the processing script from https://rentry.org/sd-e621-textual-inversion
Image data selected to exclusively feature male versions of Miles Tails Prower, as this character is popular in rule 63.
Some images edited to fit 1:1 aspect and maintain a solo focus on Miles Prower alone. Text and signatures were also removed to benefit the training process (apologies to the original artists).
Some images modified to be more character accurate such as fur color or eye color corrections.

Use by installing to an embeddings folder of your stable-diffusion-webui-master installing.
You can install multiple embeddings, if you wish to use only one you can rename it however you wish.
The more training an embed has had, the less prominent yiffy-e18's own qualities and characteristics become.
Be aware that this embedding will use 16 of your 75 allotted tokens, limiting the other tokens you may use.
Also be aware that this embedding is not effective when using with any other Stable Diffusion model: use yiffy-e18 for best results.

Baseline prompt used for X/Y Plot example:
masterpiece, best quality, by chunie and hioshiru, multi tail, e621, anthro, fox, (mmtp16:1), (detailed fluffy fur), countershading, solo male, (hyper), (knot), huge penis, long penis, orange fur, (bedroom), (shota), (young), pinup
Negative prompt: bad anatomy, bad hands, text, error, missing fingers, extra digit, fewer digits, cropped, worst quality, low quality, normal quality, jpeg artifacts, signature, watermark, username, deformed blurry bad disfigured poorly drawn mutation mutated ugly (out of focus:1.3) depth of field focal blur (extra body parts)
Steps: 50, Sampler: Euler a, CFG scale: 7.5, Seed: 1, Size: 512x512, Model hash: 50ad914b, Model: yiffy-e18

Used embeddings: hioshiru [44e7], mmtp16 [48af]
(Hioshiru embedding found at https://iwiftp.yerf.org/Furry/Software/Stable%20Diffusion%20Furry%20Finetune%20Models/ )