CivArchive
    Guyanonymous' Wildcard Prompt Studio - v3 All
    NSFW
    Preview 147724039
    Preview 147716013
    Preview 147716015
    Preview 147716135
    Preview 147716176
    Preview 147716197
    Preview 147716213
    Preview 147716214
    Preview 147716222
    Preview 147716223
    Preview 147716225

    Guyanonymous' Prompt Studio v3

    A single-file, fully offline prompt builder for wildcard files.

    I built this for my own ComfyUI / Forge Neo workflow and it grew into something worth sharing. It's one HTML file. Open it in Edge or Chrome on Windows, point it at your folder of .txt wildcard files, and design prompts with a live preview. No install, no server, no account, and it cannot upload anything (the page blocks every network request).

    Why

    Wildcard folders get huge and nested, and it's hard to know what a template will actually produce, which wildcard picked what, or whether a prompt has blown past your model's token limit. This shows all of it as you type.

    What it does

    • Scans a folder (and subfolders) of .txt wildcards, with no file limit, a filter, and resizable panels

    • Template editor with __file__ tags: autocomplete, drag-and-drop, green/red validity, and a peek button that shows a file's lines and swaps the tag for one you choose

    • Live sample with every nested wildcard resolved. Open any part to see which sub-wildcard supplied what

    • Click a part to re-roll just that part. Right-click to lock it, or pick an exact line from the file

    • Reuse one pick everywhere: ${hero=__names__} then ${hero} (great for names and text overlays), or __names$hero__

    • Optional weights (3:: text), inline choices {a|b|c}, and pick-several {2-3$$, $$a|b|c}

    • Model profiles with a token gauge (recommended length vs. hard cutoff), calibration, editable limits, and a "check length over 200 rolls" report

    • Prefix, suffix and negative prompt, and it strips (word:1.2) weights for models that ignore them

    • Saves the sample as one paragraph. Batch export of up to 1,000 random prompts (new seed each) as TXT, CSV or JSON, with split files and a matching negative file

    • Sweep export: every line of one wildcard while the rest stays random

    • "All possible prompts" export for one wildcard (reused picks counted once)

    • History, favourites, library search, health check (missing refs, loops, duplicates), where-used, tabs, autosaved versions and projects

    • Collapsible tool cards, so the right-hand column shows only what you need

    • Remembers separate folders for wildcards, templates, samples, batches and exports

    How to start

    1. Open the .html file in Edge or Chrome

    2. Click Choose folder… and pick your wildcard folder

    3. Type a prompt and use __ for suggestions

    4. Ctrl+Enter for a new sample, Ctrl+S to save the template

    The included guide covers everything in about five minutes.

    Version history

    • v3: Renamed Guyanonymous' Prompt Studio. Every right-hand card collapses to its title bar (with Collapse all / Expand all, remembered between sessions).

    • v2.5: Model profiles with a token gauge; prefix, suffix and negative prompt; TXT/CSV/JSON batch export with split files and a matching negative file; sweep export; history and favourites; library search, health check and where-used; template tabs, autosaved versions and projects.

    • v2.4: Reusable picks (${name=__a__}, ${name}, __file$name__) with a Remember… button and a values strip; right-click to lock a part or choose an exact line; weighted lines; pick-several.

    • v2.3: Drag-to-resize left panel and centre sections, wrap long file names, sizes remembered.

    • v2.2: Peek button in the template editor: see a wildcard's lines and swap the tag for one of them.

    • v2.1: Nested wildcards open up in the sample (__body__ → __head__, __torso__, __limbs__) with Expand all, Collapse all, Levels and colour by parent.

    • v2.0: Larger scrolling sample panel sharing height with the editor; no limit on the file list.

    • v1: First full rewrite: offline folder scan, live nested sample, seeded re-rolls, save/open templates, one-paragraph sample save, batch export, all-possible export, separate remembered folders.

    Good to know

    • Token counts are estimates until you calibrate them with your own tool's number, and model limits are editable starting points (some unverified), not guarantees

    • History, versions and projects live in your browser's storage. Export projects you care about

    • Use Edge or Chrome for remembered folders and saving straight into folders; other browsers fall back to Downloads

    Free to use and share. Feedback, bug reports and feature ideas welcome.

    #StableDiffusion #ComfyUI #Wildcards #PromptEngineering #Offline

    Description

    # Guyanonymous' Prompt Studio: Changelog

    From v1 to v3 (earlier versions were called Offline Wildcard Prompt Studio). Newest changes are at the bottom of each section's list; versions run oldest to newest.

    ---

    ## v1: the rewrite

    The first full rewrite of the original Gemini-generated prototype, keeping its look and headings but rebuilding how it works.

    Core

    - Single offline HTML file. A content-security rule blocks every network request, so nothing can be uploaded.

    - Scans a folder of .txt wildcard files (with subfolders) and lists them.

    - Template editor with __filename__ tags, autocomplete when you type __, and green or red highlighting for found or missing files.

    - Drag a file name into the prompt, or use the + button or double-click to insert its tag.

    - Live sample with every nested wildcard resolved to a random line, using a seeded random generator so a seed always reproduces the same sample.

    - Click a coloured part of the sample to re-roll just that part.

    - Save and open templates as .txt. Save the evaluated sample as a single paragraph with no line breaks.

    - Batch export of up to 1,000 random prompts, each with a new seed, with optional duplicate skipping and optional reproducible seeds.

    - "All possible prompts" export when the template uses exactly one wildcard, with a combination counter, a line cap and optional duplicate removal.

    - Separate remembered folders for wildcards, templates, samples, batches and all-prompts exports (saved in the browser, re-allowed on reopen).

    - Right-hand panels: output options, batch card, all-possible card, checks and a resolution trace.

    - Optional inline choices {a|b|c}, optional comment lines starting with #, tidy spacing and commas, Windows line endings option.

    - In-app wildcard file viewer and editor with Save.

    - Fallbacks for browsers without folder access (folder picker, Downloads).

    ---

    ## v2.0: version 2 and sizing

    Changed

    - Larger, scrollable sample panel. The Prompt Template editor and the Sample panel share the centre column's height equally, and both scroll when prompts get long.

    - Removed the 1,000-file display limit in the left wildcard list.

    - Labelled the app version 2 in the title and header badge.

    Fixed

    - Wildcards not appearing after loading a saved prompt. The cause was that the linked folder was the output folder rather than the wildcard folder; the guidance and checks now make that clearer.

    - A mistaken rebuild that dropped the right-hand information panels was thrown away. Version 2 was rebuilt from the file where the features worked, changing only the sizing of the centre panels.

    ---

    ## v2.1: sample breakout

    Added

    - Wildcards nested inside wildcards can open up in the Sample panel, showing the parent tag and every wildcard inside it (for example __body__ opening into __head__, __torso__ and __limbs__).

    - Each opened part shows its tag name and the text it supplied.

    - A small arrow on every part opens or closes it, and the state stays until you change it.

    - Expand all, Collapse all and a Levels control.

    - Colour by parent: parts called by the same wildcard share a colour (switchable back to colour by depth).

    ---

    ## v2.2: peek in the template editor

    Added

    - Click inside a green tag in the template editor and a ▸ button appears at the right edge of that line.

    - It opens a panel with the file's lines, a filter box, and a marker for the line used in the current sample.

    - Replace swaps that one tag for the line you choose, and Undo restores it.

    - Works for nested wildcards too: replace __body__ with a snippet that still contains tags, then keep drilling in.

    Changed

    - The button sits in a gutter at the edge of the editor so it never covers text.

    - Autocomplete no longer pops up when you click inside a finished tag.

    ---

    ## v2.3: adjustable layout

    Added

    - Drag the edge of the left panel to resize the wildcard list, which also resizes the middle column. Double-click the edge to reset.

    - Wrap long names option so long file names show in full instead of being cut off, plus tooltips on rows.

    - Drag handles between the three centre sections (Prompt template, Sample, Exactly what gets saved) to adjust their heights.

    - Sizes are remembered between sessions, can be nudged with the keyboard, and are kept within limits so nothing disappears.

    ---

    ## v2.4: reusable picks, locks and weights

    Added

    - Reusable picks. ${name=__a__ __b__} picks once and prints nothing. ${name} prints the same value everywhere. __file$name__ picks, remembers and prints in one go. Definitions can be anywhere in the template, and wildcard files can use ${name} too. $$ gives a literal dollar sign.

    - Remember… button: select text or tags in the editor, name it, and the app writes the definition and the use for you.

    - Editor colouring for variables, a red mark for undefined names, and autocomplete for ${.

    - Remembered values strip above the sample, with re-roll, lock and choose-a-line buttons for each value.

    - Right-click menu on any part of the sample: re-roll, lock, choose a specific line from a filterable list, copy text. Locks and chosen lines survive Roll new sample, with Unlock all to release them.

    - Weighted lines: 3:: text in wildcard files and inside {…}, with 0:: switching a line off.

    - Pick several: {2-3$$, $$a|b|c} and {2$$, $$__file__} choose different options and join them.

    - New checks for undefined, unused and twice-defined variables.

    - On/off checkboxes for the variable and weight syntax, so prompts with a literal ${ still work.

    - Counts and the all-possible export treat a reused value as a single choice, and every enumerated prompt keeps that value consistent.

    Fixed

    - Pressing Enter in a name prompt no longer leaks a line break into the editor.

    ---

    ## v2.5: models, exports, library and projects

    Added: model and length

    - Model profiles with an estimated token gauge under "Exactly what gets saved", showing the recommended length and hard cutoff.

    - Eighteen starter profiles across CLIP tag models, anime tag models, T5 natural-language models, LLM-encoder models, hosted models and video. Each is marked verified, estimate or unverified.

    - A Frontend setting (ComfyUI, Forge/A1111 or other) that changes how long prompts are interpreted, including 75-token chunk counts for CLIP models.

    - Manage… window: edit any profile, save it under a new name, copy or delete a profile or a whole family, import and export profiles, and revert to the originals.

    - Calibrate… to teach the counter from the token count your own tool reports.

    - Apply recommended options and Check length over 200 rolls, which reports shortest, typical and longest prompts and how many pass the limits.

    Added: output shaping and exports

    - Prefix, suffix and negative prompt fields, added to every copy, save and export, and able to contain wildcards.

    - Option to remove (word:1.2) weights for models that ignore them.

    - Batch export formats: text, CSV or JSON, with each prompt's seed and the lines that were picked.

    - Split every N prompts into numbered files, and an optional matching _negative text file.

    - A length summary after each batch.

    - Sweep one wildcard: one prompt per line of a chosen wildcard while everything else stays random, with an optional repeat count.

    Added: history and library

    - History: the last 60 samples, plus ★ Keep favourites that store their template. Restore brings back the template, seed, locks and chosen lines.

    - Library: search every line of every file, a health check (empty files, missing references, loops, duplicate and near-duplicate lines, unbalanced braces, very long lines, unused files), and where-used.

    - Remove duplicates in the file viewer.

    - Optional rescan of the wildcard folder when you return to the window, which never interrupts an unsaved edit.

    Added: tabs, versions and projects

    - Template tabs, each with its own prefix, suffix, negative prompt and model. New and Open now use tabs instead of discarding your work.

    - Versions…: autosaved and named versions with a compare view and restore.

    - Projects: save and reopen whole sets of tabs and output settings, and export or import them as files.

    - Tabs and the draft are restored when you reopen the app.

    Changed

    - Save dialogs and downloads now choose the right file type for CSV and JSON.

    - The all-possible export applies prefix, suffix and weight handling.

    - Help window updated for every new feature.

    Documents

    - A user guide and a ready-to-post share text now ship with the app.

    ---

    ## v3: renamed, collapsible panels

    Changed

    - The app is now Guyanonymous' Prompt Studio, in the page title, header, and exported JSON.

    - File renamed to guyanonymous_prompt_studio_v3.html.

    Added

    - Every card in the right-hand column collapses to its title bar when you click the title (or press Enter on it), so unused tools stay out of the way.

    - Collapse all and Expand all buttons at the top of the column.

    - Collapsed or open states are remembered between sessions.

    ---

    ## Known limits

    - Token counts are estimates until calibrated, because exact tokenizers need large vocabulary files that an offline page does not include.

    - Several model limits are starting points from research and community reports, marked estimate or unverified. Test them in your own setup.

    - History, versions, favourites, projects and tabs live in the browser's own storage. Export projects you care about, since clearing site data erases the rest.

    FAQ

    Wildcards
    Other

    Details

    Downloads
    70
    Platform
    CivitAI
    Platform Status
    Available
    Created
    11/2/2026
    Updated
    11/5/2026
    Deleted
    -

    Files

    guyanonymousWildcard_v3All.zip

    Mirrors