LoRA stack
Prompt fragment
- LoRA tags
- 0
- Trigger words
- 0
Names are cleaned up automatically — folders, the .safetensors extension and a pasted <lora:…> wrapper are stripped, and weights like 0.80 become 0.8.
A builder for the `<lora:filename:weight>` prompt syntax that Stable Diffusion UIs (AUTOMATIC1111, Forge, ComfyUI and others) use to invoke a LoRA — and it lets you stack several LoRAs at once. Give each LoRA a name, a weight (model strength) and optional trigger words per row, and it instantly produces a ready-to-paste string like `<lora:myChar:0.8>, 1girl, my character`. Names are cleaned up for you: a path keeps only its final filename, a `.safetensors` (or `.ckpt`, `.pt`) extension is stripped, and if you paste a whole `<lora:...:1>` wrapper it pulls out just the name inside. Weights are rounded to your chosen number of decimals and trailing zeros are dropped, so `0.80` becomes `0.8` and `1.00` becomes `1`. You can choose between two layouts: "LoRA + its triggers" places each LoRA's trigger words right after its tag, while "All tags, then triggers" groups every `<lora:...>` tag first and then lists the trigger words (de-duplicated) afterwards. You can also turn trigger words off to emit tags only, and set how many decimal places the weights use. The more LoRAs you stack the more tedious hand-typing becomes — and it's easy to mis-format a weight or forget to close a `<lora:...>` tag; this tool removes that friction. It's pure string assembly with no external API or data, so the LoRA names, weights and trigger words you enter are never sent anywhere — everything runs locally in your browser.
How to use
- On the first row, type the LoRA name (for example `myCharacter`, or paste `<lora:myCharacter:1>` — folders and extensions are stripped for you).
- Pick a weight from the presets (0.4–1) or type one, and add trigger words separated by commas if you want. Use "+ Add LoRA" to stack as many as you need.
- Copy the `<lora:...:...>` prompt fragment shown on the right and paste it into your generator's prompt box. Switch layout or decimals from the toolbar.
FAQ
Is anything I enter sent to a server?
No. The assembly runs entirely in your browser, with no external API or data. The LoRA names, weights, and trigger words you enter are never uploaded, stored, or sent anywhere — they are processed only on your device.
What form should the LoRA name be in?
A bare filename without extension (such as `myCharacter`) is ideal, but a path like `subfolder/myCharacter.safetensors` or a full `<lora:myCharacter:0.8>` wrapper also works. The tool strips the folder, the `.safetensors` (and similar) extension, and the `<lora:…>` wrapper, keeping just the name inside.
How are the weights formatted?
Weights are rounded to the number of decimals you set in the toolbar, then any trailing zeros and a stray decimal point are dropped. With 2 decimals, `0.80` shows as `0.8` and `1.00` as `1`. Negative weights (for cancelling effects) are allowed too.
What's the difference between the two layouts?
"LoRA + its triggers" puts each LoRA's trigger words right after its tag, so it's clear which triggers belong to which LoRA. "All tags, then triggers" lists every `<lora:...>` tag first and then the trigger words (de-duplicated) — handy when you want all the LoRA tags grouped at the start of the prompt.
Can I build just the LoRA tags without trigger words?
Yes. Turn off "Include trigger words" in the toolbar and the output is only the joined `<lora:...:...>` tags. This is useful when you keep your trigger words somewhere else.
Which tools is this syntax for?
`<lora:name:weight>` is the LoRA-invocation syntax shared by many UIs — AUTOMATIC1111, Forge, reForge, ComfyUI (with some extensions) and more. LyCORIS models use the same `<lora:...>` (or `<lyco:...>` in some setups). The fragment this tool outputs pastes straight into the prompt box.