[enhancement]: Lora Trigger Words

[Iter8tion]

Is there an existing issue for this?

• I have searched the existing issues

Contact Details

tris_tman@yahoo.com

What should this feature add?

Instead of the Lora trigger words being attached to the main Prompt Field, it would be better if the trigger prompts were automatically loaded with the loaded Lora in that Lora field. So when the lora is added down in the generation field, the trigger words are additionally added to that new Lora field. When you click the "invoke" button, the main prompt then looks to see what Loras have been added, along with trigger words, then adds them automatically and generates the the image. When you are finished with that Lora, or want to try a different Lora, then when you delete that lora, the trigger words also get deleted (because they are directly attached to that Lora). I looked around, but have not seen a direct enhancement request for this. This would be much more streamlined then having to look for the trigger words and delete them out of the main prompt window when you delete a Lora.

Alternatives

No response

Additional Content

No response

[simonfuhrmann]

This FR is over-simplifying some technical complexities and is therefore, in my opinion, infeasible. Let me give you a few examples.

Trigger words can be in different places in the main prompt, e.g., at the start, at the end, etc. Placement makes a difference. With such a UI design, this flexibility is lost.

Second, LoRAs can be trained on multiple concepts, say C1 and C2. If you want C1 in the image, you'd only include C1 and not C2. You may even choose to put C2 in the negative prompt. Again, with this UI suggestion, this is not possible.

[shanedk]

In addition, sometimes you only want the LoRA to have its effect on part of the image, in which case you put the trigger word into the Regional Guidance prompt, not in the main prompt.

There are a number of ways this would break the usability of LoRAs.

[joshistoast]

I can see a ui workaround by having a highlight on the word, and upon clicking the word the ui would prompt if they want to add the lora that trigger is associated with. This behavior could be replicated across prompt boxes except for the workflow editor (unless it could add a node).