💥 Amnesia from Majo no Tabitabi
— Amnesia’s whole journey is built around a cruel little curse: every time she falls asleep, her memories reset again. That makes her time with Elaina feel less like a simple travel companion arc and more like watching someone desperately try to hold onto herself before everything slips away. She has a gentle, sincere charm, but there is always that sad pressure behind her story because even ordinary moments can disappear overnight. For a Wandering Witch character, she brings a softer kind of tragedy instead of the usual sharp shock.
>Majo no Tabitabi COLLECTION<
⚠️Turn off your R/X/XXX Filter to see hidden JUICY Loras in >HERE<
🖼️ How I Generate My Images
I made a full article that explains how I generate my images — including my settings, prompting tips, and even my potato PC setup 😅
👉 Check it out here if you’re curious or want to try the same method.
🔞 NSFW gallery (18+)
Check my account here to see the spicy showcase that I can't show here, if any.
🎨 Prompt Tips
These are optional starting points for handling a few behaviors of this version. You may still need some retries depending on the checkpoint, prompt, and LoRA strength.
🧥 Default outfit with cloak
This LoRA is optimized for Amnesia’s default outfit with her cloak, so keeping the cloak in the prompt usually gives the most stable and canon-like result.
Recommended:
✅ Positive prompt: white_cloak, hooded_cloak, hood_down
You can try removing the cloak or coat, but the design may become less accurate compared to her normal look.
If you want to reduce the cloak/coat:
🚫 Negative prompt: white_cloak, hooded_cloak, hood_down, multicolored_coat, long_sleeves, two-sided_coat, tailcoat, white_belt
You may also need to lower the LoRA weight a bit if the cloak keeps coming back.
🔄 Back-view cloak prompt
Be careful when generating her from behind. In the anime, her cloak covers almost everything from the back, and only her feet up to around the ankle area are visible.
For a cleaner back-view result, try:
✅ Positive prompt: from_behind, standing, white_cloak, hooded_cloak, hood_down
For this angle, avoid forcing too many visible body or outfit details. Her default outfit is mostly hidden from behind, so adding shirt, dress, skirt, or leg-focused tags may make the cloak silhouette less accurate. You can still experiment with other outfit tags, but because the dataset has limited back-view variation, non-canon outfit combinations from this angle may need more retries.
📌 Limited dataset variety
Since she has limited screen time in the main anime, some uncommon poses, angles, or outfit changes may need retries.
For better stability:
💪 LoRA weight: 0.8–1
For outfit flexibility:
💪 LoRA weight: 0.5–0.8
Lowering the LoRA weight can help when changing her outfit, but going too low may also weaken her canon face/outfit identity.
🧥 Default outfit details appearing on other outfits
When a default outfit detail keeps appearing in a different outfit, add the exact leaking detail to the negative prompt.
For example, if the two-sided coat details appearing when you do not want it:
🚫 Negative prompt: two-sided_coat
Use the specific leaking tag instead of adding many unrelated negative tags.
⚖️ LoRA strength and outfit flexibility
A weight of 1 can still work with a very different outfit, especially when the new outfit is clearly different from her default one.
Only lower the LoRA weight gradually when outfit details keep leaking even after using clear outfit tags and negative prompts. Lowering it too much may also weaken her face, hair, and overall character identity
📊 Lora Project
If you want to know what characters that I'm covering next you can check these link :
👉 Personal Project
👉 Commision Project
If this model sparked something in you, show some love with a like, review, or buzz.
Now go—
Be wild
Description
🎬 Main anime source only
This LoRA was trained mainly using screenshots from the main Majo no Tabitabi anime series, with a few official art references added for extra support.
I did not use the movie version for this training. Since Amnesia barely appears in the main anime, the dataset variety is limited, so some poses, angles, and outfit flexibility may not be as strong as characters with more screen time.🧥 Optimized for default outfit + cloak
This version is mainly optimized for her default outfit with the cloak. You can still try prompting her without the cloak or without the coat, but the result may not look as clean or as close to her canon design.
For the most stable result, I recommend keeping the cloak when generating her default look.🔄 Be careful with back-view prompts
Her cloak has a specific silhouette from behind. In the anime, when she is seen from the back, the cloak covers most of her body and only her feet up to around the ankle area are visible.
Because of that, some body/leg visibility tags may fight the canon look when prompting from_behind. If you want a more accurate back-view result, avoid forcing too much visible leg detail.As usual, the result can still depend on your checkpoint, prompt, LoRA strength, and generation settings.
FAQ
Comments (4)
Hellow
I'm using "Forege".
Your LoRA uses a double underscore in the "output name".
"Dynamic Prompts (Wildcard)" will throw an error if you use a double underscore.
A single underscore will not cause an error.
Example
__:NG
_:OK
@whitewing2012 Hmm..., I think it's because my LoRA post name has "|" in it. Since "|" is not a valid character for filename, the system automatically remove it, so instead of "aaa_|_bbb" it give "aaa__bbb", resulting in double underscore. I think as long as you rename the downloaded file, it will be fine.
So, since you enabling your wildcard feature, it will think that my lora has incomplete wildcard, since it start with "__" but there are no "__" at the end, resulting in errors that you experiencing.
I don't really have plan to change my way of naming my LoRA post honestly. Need to think about it through.
回答有難うございます
ComfyUIでは、loraはファイル名称でコールします。
A1111やForegeではLora生成時に設定する”Output name”で決まるみたいです。
今回の設定では、以下となっているので、Dynamic Prompts(wildcard)でエラーとなってしまいました。
Output name: Amnesia__Majo_no_Tabitabi
私のエラー回避として、ヘッダー情報を書き換えるということをしています。
このページにloraファイルをupすると、生成情報が見れるので、"Output name"の名称を"Amnesia_Majo_no_Tabitabi"に書き換え保存しています。
<lora:Amnesia__Majo_no_Tabitabi:1> -> <lora:Amnesia_Majo_no_Tabitabi:1>
**********************************************************************************
Thank you for your reply.
In ComfyUI, lora is called by its file name.
In A1111 and Forege, it seems to be determined by the "Output name" set when generating the lora.
In this case, the setting is as follows, which caused an error in Dynamic Prompts (wildcard).
Output name: Amnesia__Majo_no_Tabitabi
To avoid my error, I'm rewriting the header information.
When I upload a lora file to this page, I can see the generation information, so I rewrite the "Output name" to "Amnesia_Majo_no_Tabitabi" and save it.
<lora:Amnesia__Majo_no_Tabitabi:1> -> <lora:Amnesia_Majo_no_Tabitabi:1>
@whitewing2012 I guess you confirming back to me about what happen to you and what method you are using to solve this problem, so.. thanks for that.
I've seen you generate something so I'm assuming everything works now.
After read your reply, I think I know the field that I need to carefully type in. I'll try something next time to minimize this situation.
教えてくださってありがとうございます。