#Kohya_ss
LightNews LightNews
handle.invalid
完了したら通知音を鳴らしてお知らせしてもらえたら嬉しくないですか？
#zenn #kohya_ss
zenn.dev/neon_aiart/a...
kohya_ssに完了時に通知音（アルペジオ）を鳴らすコードを追加したお話
zenn.dev
July 3, 2026 at 3:47 PM
kandagi.bsky.social
kohya_ssのパラメーターを確認中だけど、なるほどね？
完全に理解したわ。
だが今日はこの辺にしておいてやろうヾ(⌒Ｕ˙ꈊ˙)ｵﾎﾞｴﾃﾛﾖｫ
#AIイラスト #AIart
November 5, 2025 at 11:55 AM
theophite.bsky.social
i'm using modified kohya_ss scripts remotely and OneTrainer locally.
February 14, 2025 at 5:48 AM
matiere-lab.bsky.social
【質感LoRAを題材とした、オリジナルLoRA開発コースのご紹介】
texture-lora-lab.com

#AIart️️️️️️️️️️️️️️️️️️️️️️️️️️️️️️️️️️️️️️️️️️️️️️️️️️️ #AIイラスト #質感LoRA #A1111 #Kohya_ss

※こちらのコースはRunpod環境で生成しておりますが、#Floyo 環境でもできるよう現在調整いただいております。続報は6月予定です。
May 17, 2026 at 9:07 AM
theophite.bsky.social
i can walk you through setting an instance up. i use modified kohya_ss scripts for training but OneTrainer is actually easier and more flexible.
November 10, 2024 at 1:43 AM
zentrocdot.bsky.social
I updated by installation guide for kohya_ss.

#AI #AIart #AIartwork #AIimage #AIimagegeneration #AIgenerated #digitalart #digitalartwork #SDXL #LoRA #kohya_ss #training

www.copus.io/work/89cbb49...
Installation of kohya_ss on Linux for SDXL Trainings
Installation of kohya_ss on a clean SOTA Linux system
www.copus.io
February 13, 2026 at 10:18 AM
ai.hydros.tv
Mirko from My Hero Academia 💖
Mirko (My Hero Academia) AI Image Generated using Stable Diffusion AUTOMATIC1111 Character LoRA Trained in Kohya_ss
November 20, 2023 at 4:59 AM
theophite.bsky.social
locally: 24 gig 4090

remote: docker container on Runpod, 80 gig A1000

training: kohya_ss remote, OneTrainer local

model merge / VAE & text-encoder manipulation: A1111 extensions

inference: ComfyUI
September 15, 2024 at 5:25 PM
keiwa-shohan-enei.bsky.social
備忘録を投稿しました。
「LoRA Dataset Assistantを開発」
LoRA学習前のタグ付けから確認までを一つのツールで。
#備忘録
#LoRADatasetAssistant #LoRA学習 #WD14 #kohya_ss
LoRA Dataset Assistantを開発
560days.com
October 4, 2026 at 10:43 AM
theophite.bsky.social
largely using modified kohya_ss scripts. comfyUI with some custom nodes, using SDXL as the base model. probably can't upgrade because i need to be able to tune the text models to do what i'm doing.
January 11, 2025 at 11:21 AM
ai.hydros.tv
Nefertari Vivi from One Piece💖
Image generated using Automatic1111 LoRA trained in Kohya_ss
November 23, 2023 at 7:26 AM
matiere-lab.bsky.social
【FloyoでLoRA学習】2時間以上かかっていた学習が44分になった！ — Kohya_ssの設定をH100に持っていった全記録
note.com/texture_lora_l…#Floyoy#ComfyUIU#kohya_sss#LoRAR#質感LoRARA
https://note.com/texture_lora_l…
July 16, 2026 at 8:56 PM
zentrocdot.bsky.social
The squirrel at its best.

#AI #AIart #AIartwork #AIimage #AIimagegeneration #AIgenerated #digitalart #digitalartwork #promptart #GenerativeAI #workflow #txt2img #ComfyUI #LoRA #kohya_ss
March 24, 2025 at 2:20 PM
kandagi.bsky.social
ねこ　吸いますか？

おはようございます
大まかなパラメーターが決まったので、今日から本格的にKohya_ssを動かしてLora作ってみようと思います。
んで、そのうち自分専用のモデルとかも作ってみたいね、とも考えてます。

それでは今日も一日頑張って行きませうヾ(⌒Ｕ˙ꈊ˙)ｻﾑﾆｬﾝ
November 5, 2025 at 11:19 PM
zentrocdot.bsky.social
Now I am able to train Embeddings again. This works now well using kohya_ss. Now I can test in SDXL what I ever wanted to do.

#AI #AIart #AIartwork #AIimage #AIimagegeneration #AIgenerated #digitalart #digitalartwork #cat #embedding #SDXL
February 17, 2026 at 8:34 AM
zentrocdot.bsky.social
The LoRA model now works for almost everything.

#AI #AIart #AIartwork #AIimage #AIimagegeneration #AIgenerated #digitalart #digitalartwork #promptart #GenerativeAI #workflow #txt2img #ComfyUI #LoRA #kohya_ss
March 24, 2025 at 2:25 PM
zentrocdot.bsky.social
The LoRA is moving in the right direction.

The garlic LoRA model does not yet want to succeed.

#AI #AIart #AIartwork #AIimage #AIimagegeneration #AIgenerated #digitalart #digitalartwork #promptart #diffusionmodel #GenerativeAI #workflow #txt2img #workflow #ComfyUI #LoRA #kohya_ss #garlic #dragon
March 23, 2025 at 8:03 AM
theophite.bsky.social
yeah, your starter scripts are basically going to be kohya_ss or OneTrainer, and then you're going to want to read Peli Grietzer's A Theory of Vibe. what you are doing is more or less straightforward finetuning, but _conteptually_ what you are doing is DPO with the text encoder as the reward model.
November 27, 2024 at 8:49 PM
zentrocdot.bsky.social
The new version of kohya_ss is a little bit strange. It took a while to get Textual Inversion training running. Here comes the first result.

#AI #AIart #AIartwork #AIimage #AIimagegeneration #AIgenerated #digitalart #digitalartwork #cat #embedding #SDXL
February 16, 2026 at 7:55 PM
zentrocdot.bsky.social
The garlic LoRA model does not yet want to succeed.

#AI #AIart #AIartwork #AIimage #AIimagegeneration #AIgenerated #digitalart #digitalartwork #promptart #diffusionmodel #GenerativeAI #workflow #txt2img #workflow #ComfyUI #LoRA #kohya_ss #garlic #cat
March 23, 2025 at 7:46 AM
zentrocdot.bsky.social
Using brackets in Prompts effects the usability of the LoRA. (telephone box) is not made of balls.

#AI #AIart #AIartwork #AIimage #AIimagegeneration #AIgenerated #digitalart #digitalartwork #promptart #GenerativeAI #workflow #txt2img #ComfyUI #LoRA #kohya_ss
March 24, 2025 at 1:54 PM
pixelgenesis.bsky.social
Vibe-coded the entire AI art workflow with @anthropic.com's Claude Code

• Everything runs locally
• Used 37 of my hand-drawn sketches as training data
• Kohya_ss to train a custom LoRA on SD1.5 base model
• LoRA learned my unique sketch style
• ComfyUI loads the LoRA to generate new art
comfyUI setup
August 13, 2025 at 7:57 AM
menofnearbush.bsky.social
kohya_ss was giving me errors but chatgpt noticed my txt caption files (that were generated by that comfyui node) needed to be converted to UTP-8 😒 and gave me a command line to run on the folder
November 26, 2024 at 1:38 AM
zentrocdot.bsky.social
I played around with the kohya_ss settings. Nevertheless the resulting test LoRA is working.

civitai.com/images/64657...

#AI #AIart #AIartwork #AIimage #AIimagegeneration #AIgenerated #digitalart #digitalartwork #promptart #diffusionmodel #GenerativeAI #workflow #ComfyUI #cat #LoRA #kohya_ss
Image posted by zentrocdot
civitai.com
March 20, 2025 at 4:15 PM
zentrocdot.bsky.social
First failed test of a new garlic LoRA for a tutorial. Quality is not as good as it should be.

#AI #AIart #AIartwork #AIimage #AIimagegeneration #AIgenerated #digitalart #digitalartwork #promptart #diffusionmodel #GenerativeAI #workflow #txt2img #workflow #ComfyUI #LoRA #kohya_ss #garlic #cat
March 23, 2025 at 6:59 AM