I introduced mechanism of Textual Inversion in this (↓) article.
>-
Since I learned mechanism, this time I will introduce method to actually use embedding using ComfyUI. Furthermore, through generation results, I will verify its effect and compare difference in generation results.
Work Flow
1. Introduction of ComfyUI
Please refer to this article.
>-
2. Download favorite embedding
Download from favorite place and place in ComfyUI/models/embeddings. (Types are few but direct installation from ComfyUI-Manager UI is also possible)
This time I borrowed embedding that makes output Goblin style from ↓.
This is a Style embedding which makes your subject go full goblinmode, literally! Think green skin, horns and long ears! Simply throw Style-Goblinm...
3. Choose Workflow
Since I only tweak prompt this time, any workflow is fine.
4. Generate
Generate by specifying embedding: {embedding_name} in prompt.
- In my case since filename was
Style-Goblinmode.pt, I enteredembedding: Style-Goblinmodein positive prompt.
Generation Result
Without embedding
dark sky, rain, midnight,distressed texture, a portrait of a woman, challenging face, cyberpunk city,confident pose,
With embedding
dark sky, rain, midnight,distressed texture, a portrait of a woman, challenging face, cyberpunk city,confident pose, embedding:Style-Goblinmode,
Conclusion
Since I entered from content of paper this time, I explained by method of putting embedding in positive prompt.
But, actually embedding is more famous for method putting in negative prompt than method putting in positive prompt. Speaking of famous ones, embeddings like EasyNegative (Is it EasyNegativeV2 now?) are often used, and it is known that quality of picture improves considerably by putting this in negative prompt. It can be used quite easily so please try that too if you feel like it.
If you want to try other techniques related to image generation, please utilize the link collection below.
>-
