Prompt
Converted
A converter that moves the emphasis (weight) notation in a Stable Diffusion prompt between tools without breaking it. Pick a source and target format and it bulk-rewrites Automatic1111 / WebUI syntax like `((best quality))` and `[blurry]`, NovelAI's `{{best quality}}`, and explicit weights like `(best quality:1.4)`. So when you carry a prompt written in NovelAI over to A1111, you no longer have to hand-edit `{{}}` into `(())`, or expand stacked brackets into a numeric `:1.3` weight, by hand. The output updates in real time as you type and shows how many weighted terms it found. Because each notation uses a different multiplier per bracket level (A1111 ≈ 1.1×, NovelAI ≈ 1.05×), values that can't be expressed in whole bracket levels are rounded to the nearest level, and the tool clearly flags when a conversion is approximate. It's pure string processing with no external API or dictionary, so your work-in-progress prompt is never sent anywhere — your prompt is the blueprint of your work, so this tool makes no network calls and does all conversion locally in your browser.
How to use
- Paste your prompt into the input box — it's fine if emphasis like `((...))`, `{{...}}`, or `(...:1.3)` is mixed in.
- Pick the "From" and "To" formats (A1111 / NovelAI / explicit weight). Use "Swap" to flip the direction if needed.
- Hit "Copy" on the converted output and paste it straight into the prompt box of the tool you want to use.
FAQ
Is my prompt sent to a server?
No. All conversion is plain string processing in your browser, with no external API or dictionary. The prompt you enter is never uploaded, stored, or sent anywhere — it is processed only on your device.
How do A1111 and NovelAI brackets differ?
Both strengthen emphasis as you stack brackets, but the multiplier per level differs. Automatic1111 / WebUI treats round brackets `()` as ≈1.1× and square brackets `[]` as weaker, while NovelAI treats curly braces `{}` as ≈1.05× and `[]` as weaker. The tool accounts for this when converting.
How does conversion to explicit (word:1.2) weights work?
It expands bracket levels into a numeric weight (e.g. `((word))` → `(word:1.21)`), or folds a numeric weight back into the nearest number of brackets. Numbers that can't be expressed exactly in whole bracket levels become approximate, and that is flagged below the output.
Why does it sometimes say "approximate"?
When you move, say, NovelAI's `{}` (≈1.05×) to A1111's `()` (≈1.1×), the differing per-level multiplier means the original weight can't always be reproduced exactly with whole brackets. In that case the tool rounds to the nearest level and flags it as approximate. If you need an exact value, explicit (word:1.2) weights are the reliable choice.
Are square-bracket [ ] de-emphasis handled too?
Yes. In both A1111 and NovelAI, `[word]` is treated as de-emphasis and mapped to the target notation. In explicit weights it is expressed as a weight below 1.0 (e.g. `(word:0.9)`).
I only want to change weights — can it also format the tags?
This tool focuses on converting weight notation. For formatting the tags themselves — underscores ⇄ spaces, escaping parentheses, removing duplicates — pair it with the tag-format tool.