TagGUI Plus

TagGUI Plus is a cross-platform desktop app for quickly adding and editing image tags and captions. It is aimed at people building image datasets for generative AI models, and it focuses on getting through a folder of images fast with a keyboard-friendly workflow, autocomplete, and built-in automatic captioning.

About this fork

This is a personal fork of jhc13/taggui. Everything beyond the original app by jhc13 was entirely "vibe coded" with GitHub Copilot for my own personal use. I'm sharing it in case others find it useful, but because of how it was built there are no guarantees that this fork will be maintained or updated regularly, and there may be rough edges. Use it at your own discretion.

Contents

• Features
• Installation
• Uninstalling / removing
• Getting started
• The panes
• The tools
• Automatic captioning
• Advanced syntax reference
• Controls and context menus

Features

• Fast, keyboard-friendly tagging — move through a folder and add, edit, reorder, and delete tags without reaching for the mouse.
• Tag autocomplete built from your own local tag library, so suggestions follow you across every folder.
• Tag Library with categories (custom names and colors), aliases, and implications for consistent, structured tagging.
• Natural language mode — store freeform prompt text alongside tags in the same caption file.
• Automatic captioning and tagging using a range of local vision models, including Qwen3-VL caption models and the PixAI v0.9 tagger.
• Batch tag operations — find and replace, sort, shuffle, reverse, and reorder tags across many images at once.
• Advanced filtering for both images and tags, with prefixes, wildcards, numeric comparisons, and AND / OR / NOT logic.
• Completion tracking — mark images as done to keep your place, with an optional check icon on finished thumbnails.
• Integrated token counter for estimating caption length.
• Directory analytics, duplicate image detection, and booru wiki lookups built in.
• Import / export settings and full keyboard-shortcut customization.

Installation

This fork does not publish prebuilt releases, so you install it by running it from source. The steps below are written to be followed even if you have never used Python or PowerShell before.

1. Install Python

Install Python 3.12 (recommended; 3.11 also works) from python.org.

• On Windows: during installation, tick the box that says "Add Python to PATH" before clicking Install. This lets you run python from any terminal.

2. Get the code

Download this repository to your computer. The simplest way is:

1. Click the green Code button at the top of the GitHub page.
2. Choose Download ZIP.
3. Extract the ZIP somewhere you want to keep the app (for example your Documents folder).

(If you are comfortable with Git you can instead git clone the repository.)

3. Set up a virtual environment and install dependencies

A "virtual environment" is a private folder that holds this app's Python libraries so they don't interfere with anything else on your computer. TagGUI Plus's launcher automatically uses one named venv or .venv if it sits next to start.py.

On Windows, open PowerShell, then run these commands one at a time. Replace the path in the first command with wherever you extracted the app:

# 1. Go into the app folder (change this path to match your computer)
cd "C:\Users\YourName\Documents\taggui"

# 2. Create a virtual environment named "venv"
python -m venv venv

# 3. Activate it (your prompt will now start with "(venv)")
.\venv\Scripts\Activate.ps1

# 4. Install all the required libraries (this can take a while and downloads
#    several gigabytes the first time)
pip install -r requirements.txt

If step 3 fails with a message about scripts being disabled, run Set-ExecutionPolicy -Scope CurrentUser RemoteSigned once, answer Y, then try the activate command again.

On macOS / Linux, open a terminal and run:

cd /path/to/taggui
python3 -m venv venv
source venv/bin/activate
pip install -r requirements.txt

4. Run the app

From the app folder, run:

python start.py

start.py automatically finds and uses the venv you created, so you don't have to activate it every time. On Windows you can also just double-click start.bat, which does the same thing.

GPU note: Automatic captioning runs much faster on a compatible NVIDIA GPU, but CPU generation also works. On Linux you may also need to install the system package libxcb-cursor0 (see this Stack Overflow answer).

Uninstalling / removing

TagGUI Plus stores some data outside its own folder, so deleting the app folder by itself does not remove everything. Your caption .txt files always live next to your images and are never touched by uninstalling.

Tip

Planning to reinstall later? Export your settings first (File → Export / Import Settings…) and keep the exported file somewhere safe. Removing app data clears your saved preferences, Tag Library, and other settings, so exporting first lets you import everything back after you reinstall instead of setting it all up again from scratch.

The easy way (recommended)

Use the built-in cleanup tool: File → Settings → Remove App Data…. Tick what you want to delete — caches and settings are selected by default, and downloaded models are optional and off by default. TagGUI Plus deletes the selected items and then closes immediately so nothing gets written back. Afterwards, delete the app folder yourself (the app can't delete its own folder while it's running).

You can also clear just the thumbnail cache from Settings → Clear Thumbnail Cache.

What gets stored where

Full list of data locations (click to expand)

What it is	Windows	macOS	Linux
App settings (preferences, autocomplete/local tags, categories, hidden models, etc.)	Registry key HKEY_CURRENT_USER\Software\taggui\taggui	~/Library/Preferences/com.taggui.taggui.plist	~/.config/taggui/taggui.conf
Thumbnail cache	%LOCALAPPDATA%\taggui\thumbnails	~/Library/Caches/taggui/thumbnails	~/.cache/taggui/thumbnails
Scan caches (dimension_cache.json, caption_cache.json, completion_cache.json)	%LOCALAPPDATA%\taggui	~/Library/Caches/taggui	~/.cache/taggui
Downloaded captioning models (often several GB)	Your configured models directory, or the Hugging Face cache %USERPROFILE%\.cache\huggingface (or wherever HF_HOME points)	Your models directory, or ~/.cache/huggingface	Your models directory, or ~/.cache/huggingface

Removing the leftover data by hand (click to expand)

Windows (run in PowerShell, one line at a time):

# Thumbnail cache and scan caches
Remove-Item -Recurse -Force "$env:LOCALAPPDATA\taggui"

# Saved settings (stored in the Windows Registry, not a file)
Remove-Item -Recurse -Force "HKCU:\Software\taggui"

# Downloaded models in the shared Hugging Face cache.
# Only run this if no other tool on your system uses Hugging Face models,
# since this cache is shared. If you set a custom models directory in
# Settings instead, delete that folder rather than this one.
Remove-Item -Recurse -Force "$env:USERPROFILE\.cache\huggingface"

macOS / Linux (run in a terminal):

# Caches (thumbnails + scan caches)
rm -rf ~/.cache/taggui          # Linux
rm -rf ~/Library/Caches/taggui  # macOS

# Saved settings
rm -f ~/.config/taggui/taggui.conf                     # Linux
rm -f ~/Library/Preferences/com.taggui.taggui.plist    # macOS

# Downloaded models in the shared Hugging Face cache (see the warning above)
rm -rf ~/.cache/huggingface

Finally, delete the TagGUI Plus app folder itself.

Getting started

Loading a directory

Click Load Directory in the center of the window (or File → Load Directory) and pick the folder that contains your images. TagGUI Plus shows every image in that folder (and its subfolders) as a scrollable list of thumbnails.

Note

Background thumbnail caching. Unlike the original TagGUI, this version builds thumbnails for a directory in the background as soon as it loads, instead of only when an image scrolls into view. It saves those thumbnails to disk so the work only happens once. This means the first time you load a large directory it may feel a little slower while the cache fills in (this happens quietly in the background and won't block you from working), but every load after that is fast because the thumbnails are already stored. You can turn this off or clear the cache under File → Settings → Thumbnail cache if you prefer.

Note

Automatic refresh on refocus. When you switch back to TagGUI Plus after changing files elsewhere — for example editing an image in an external image editor — it automatically detects what changed on disk and updates the affected thumbnails, captions, and file list for you. This refresh is incremental: it only re-reads the specific files that actually changed rather than reloading and re-scanning the entire directory every time, so it stays fast even for very large folders. You don't need to manually reload the directory; just click back into the TagGUI Plus window and the changes appear.

How tags and natural language are stored

TagGUI Plus reads and writes a plain-text .txt file next to each image, with the same name as the image (for example cat.png → cat.txt). You never edit these files directly — TagGUI Plus saves your changes automatically as you work.

Each caption file has two parts:

• Tags — the first line, a list of short terms separated by your tag separator (a comma by default). This is what most tag-based training tools expect.
• Natural language — any text after the first line break. This is freeform prompt text and is optional.

So a caption file might look like this:

1girl, red dress, smiling, outdoors
A young woman in a red dress smiles while standing in a sunny garden.

Autocomplete suggestions are a separate thing: they are stored locally by TagGUI Plus (not in the caption files), so the tags you use build up a personal library that follows you across every folder. You manage that library in the Tag Library (see below).

Adding tags and natural language

• Add a tag: type it into the Add Tag box in the Image Tags pane and press Enter. Press Ctrl+Enter to accept the first autocomplete suggestion.
• Add a tag to many images at once: select multiple images in the list first, then add the tag.
• Add natural language: turn on Natural language mode in the Image Tags pane and type into the text box.

The panes

TagGUI Plus's window is made of dockable panes around a central image preview. You can show or hide each one from the View menu, and drag them to rearrange or float them.

Images pane

The scrollable list of thumbnails for the loaded folder.

• Filter Images box at the top narrows the list using the image filter syntax (tags, names, paths, counts, and logic).
• Sort the list by path, name, modified/created date, file size, resolution, tag count, token count, or natural-language length.
• Thumbnails can show a resolution badge, a green completion check, and an [NL] marker for images that have natural language text. Tags shown under a thumbnail are colored by their category.
• Select multiple images with Ctrl/Shift+click to edit or caption them as a batch.

Image preview (center)

The large view of the currently selected image.

• Zoom with the mouse wheel or Ctrl++ / Ctrl+-.
• Pan a zoomed image by dragging, or with Alt+arrow keys.
• Reset the view by double-clicking or pressing Ctrl+0.

Image Tags pane

Where you edit the current image's caption.

• Add Tag box with autocomplete for adding tags.
• Tag list below it — double-click to rename, drag to reorder, Delete to remove. Right-click for options like copying, looking a tag up on a wiki, or assigning it to a category.
• Natural language mode button toggles the pane between the tag list and a freeform text editor for the natural-language part of the caption.
• A token counter shows the caption length and turns red when it passes the token limit. A green Complete label appears when the image is marked as complete.

All Tags pane

A live list of every tag used across the loaded folder, with how many times each is used.

• Search Tags box filters the list with the tag filter syntax (name, category, count, length, wildcards, logic).
• Click action toggles what selecting a tag does: filter the image list for that tag, or add the tag to the selected images.
• When filtering by several tags at once, choose AND or OR logic.
• Sort by frequency, name, or category, ascending or descending.
• Right-click a tag to copy it, look it up on a wiki, assign/clear its category, or rename/delete every instance of it.

Auto-Captioner pane

Configures and runs automatic captioning. It holds the model selector, prompt box, destination and position options, and the Start Auto-Captioning button, plus a progress bar and console output while it runs.

The tools

Most tools live in the Tools and Edit menus.

Tag Library

Tools → Tag Library… (Ctrl+L). Your personal, structured tag vocabulary — the source of autocomplete and category coloring.

• Tags tab: add, rename, remove, and search tags; assign each a category. Import and export tags (CSV / Excel), or download starter templates.
• Categories tab: create categories with custom names and colors.
• Aliases tab: map alternate spellings to a canonical tag.
• Implications tab: when you add one tag, related tags can be added automatically (controlled by a setting).
• Profiles tab: build named sets of tag mappings (original tag → replacement). Apply a profile to swap every matching tag across the loaded directory in one step, or revert it to change those tags back — handy for switching a folder between different tag vocabularies or naming conventions.

Find and Replace

Edit → Find and Replace… (Ctrl+R). Batch-replace text in tags and captions.

• Scope: current image, all selected images, or the entire folder.
• Whole tags only matches complete tags as units (leave the replace box empty to delete matching tags, or fill it to rename them).
• Use regex for find text enables regular-expression patterns.
• The button shows a live count of how many instances will change.

Batch Reorder Tags

Edit → Batch Reorder Tags… (Ctrl+B). Reorder tags across all selected images in one action:

• Sort alphabetically, by frequency, or by category order.
• Reverse or shuffle the tag order.
• Move specific tags to the front (comma-separated).
• Optionally keep the first tag fixed during any operation.
• For category sorting, drag the categories into the priority order you want.

You can pick a default reorder in Settings so it's preselected here and can be triggered directly with the Apply Default Batch Reorder shortcut.

Find Duplicates

Tools → Find Duplicates…. Detects identical and near-identical images using perceptual hashing and shows each group side by side.

• A strictness slider controls how similar images must be to count as duplicates (0 = visually identical, higher = more lenient).
• For each group you keep one image; the rest can be sent to the Recycle Bin/Trash, moved to a folder, or tagged as duplicate for later review.

Warning

The strictness slider is very sensitive to variant sets — groups of images made up of a base image plus one or more alternate versions (edits, recolors, minor touch-ups, etc.). Even a slightly lenient setting can flag these intentional variants as duplicates. If your dataset contains a lot of variant sets, it's strongly recommended to keep the strictness very strict, or simply leave it on exact / identical only (0), to avoid grouping images you meant to keep separate.

Directory Analytics

Tools → Directory Analytics…. A read-only report on the loaded folder — nothing is changed. Tabs cover an overview (counts, size, formats), resolution stats, captions & tags coverage, housekeeping (orphaned or missing caption files), and a per-subfolder breakdown. Results can be exported to CSV or Markdown.

Danbooru / Gelbooru Wiki

Tools → Danbooru Wiki… (Ctrl+D) and Tools → Gelbooru Wiki… (Ctrl+G), also available by right-clicking a tag. These look up a tag's meaning on the Danbooru or Gelbooru wikis so you can learn what a tag means and see related tags, then add it to your Tag Library or to the selected images.

The Danbooru Wiki in particular does more than a plain tag lookup:

• Wiki lookup: read a tag's description and see its related and aliased tags.
• Tag groups: browse Danbooru's curated tag-group pages, which collect related tags together by theme — a handy way to discover tags you might not have thought to search for.
• Post search: view example posts for a tag directly inside the dialog. This is meant to give you extra visual context for what a tag actually represents, not to serve as a replacement for browsing posts on the Danbooru site itself — it's a quick reference, not a full booru browser.

You reach tag groups and post search through the search bar using simple prefixes:

• To search a tag group, type tag group: followed by the group name — for example tag group:hair. Typing just tag group: on its own suggests matching groups via autocomplete, but for a complete list it's better to look up the tag groups wiki page itself (type tag groups into the search bar with no prefix), which lists every tag group in its table of contents.
• To search posts for a tag, type posts: followed by the tag — for example posts:no humans.
• Typing anything without a prefix does a normal wiki lookup for that tag.

Note

These wiki tools depend on Danbooru's and Gelbooru's external websites and APIs. If those sites change how their data is served (or restrict access), the lookups may stop working until TagGUI Plus is updated to match. In other words, this feature relies on services outside of TagGUI Plus's control and could break unexpectedly.

⚠️ NSFW warning: Danbooru and Gelbooru are adult (NSFW) image board sites. Because these tools query those sites directly, the wiki lookups in TagGUI Plus are NSFW — search results and related content can be explicit. Use them with that in mind.

Settings

File → Settings… (Ctrl+Alt+S) collects all preferences.

Overview of the available settings (click to expand)

• Global: font size, light/dark theme, and a button to customize keyboard shortcuts.
• Images pane: list text size, which file types to show, thumbnail width, the resolution badge (and its size/transparency), the completion check icon, max preview zoom, and auto-focusing the Add Tag box when you type. Note: setting the max preview zoom very high can cause performance issues when zoomed in on large images. TagGUI Plus mitigates this — for example it automatically switches to faster (lower-quality) rendering above 10x zoom and sharpens the image once you stop zooming — but it's still recommended to keep this value on the lower side.
• Thumbnail cache: max cache size, background caching, and a button to clear the cache. If you work with very large directories, consider raising the max cache size so more thumbnails stay cached on disk instead of being evicted and regenerated later.
• Image Tags pane: the tag separator character, whether to add a space after it, autocomplete on/off, the token limit, and the default batch-reorder operation.
• Auto-Captioner: a custom models directory and per-model visibility (hide models you don't use).
• Tag Library: default keep/remove behavior, default category for new tags, whether to prompt for a category, and how implications are auto-applied. Tip: if you'd rather ignore the Tag Library and have an experience closer to the original TagGUI, set Default choice: keep or remove from Tag Library to Remove, set Default category for new tags to No category, and uncheck both Ask whether to keep or remove from Tag Library and Ask before assigning category to new tags.
• External tools: the image editor executable used by "Open Image in Configured Editor".
• Footer: Remove App Data… and Reset to Defaults.

You can also export and import your settings (File → Export / Import Settings…), choosing which parts (Tag Library, auto-captioner settings, completion marks) to include.

Automatic captioning

Beyond manual tagging, TagGUI Plus can generate captions or tags for your images locally using vision models.

To use it: select the images you want in the Images pane, choose a model in the Auto-Captioner pane, then click Start Auto-Captioning. The first use of a model downloads and loads it (this can take a few minutes and several GB); later runs are much faster. If you keep previously downloaded models in a local folder, set that folder in Settings to include them in the model list. Generated output can go to the tag list or the natural language section of the caption.

Supported models

TagGUI Plus supports two kinds of models:

• Caption models produce natural-language descriptions: Florence-2, JoyCaption, Moondream, Phi-3-Vision, Kosmos-2, LLaVA / LLaVA-Next / LLaVA-Llama-3, and the Qwen-VL family — including Qwen2-VL, Qwen2.5-VL, and Qwen3-VL (with both Instruct and Thinking variants).
• Tagger models produce booru-style tags: the WD (Waifu Diffusion) Tagger family and the PixAI Tagger — including PixAI v0.9 (deepghs/pixai-tagger-v0.9-onnx).

Qwen3-VL requires a recent Transformers version, which the pinned requirements.txt already provides.

Captioning options

Prompt, destination, and generation options (click to expand)

• Prompt — instructions for the model. You can insert per-image information with template variables: {tags} (the image's tags), {name} (file name without extension), and {directory} / {folder} (containing folder name). Example: Describe the image using these tags: {tags}.
• Start caption with — text every generated caption begins with.
• Destination — send output to Tags or Natural language (saved per model).
• Tag position / Natural language position — where generated text is inserted relative to existing content.
• Remove tag separators in caption — strip commas from generated captions before adding them as tags.
• Discourage from caption — comma-separated words/phrases to avoid (e.g. appears,seems,possibly). Not guaranteed due to tokenization.
• Include in caption — comma-separated words/phrases that should appear; use | to let the model pick one from a group (e.g. cat,orange|white|black).
• Standard generation parameters (beams, sampling, temperature, top-k/top-p, repetition penalty, token limits, etc.) are also available and follow the Hugging Face generation docs.

Tagger options and tag-filter rules (WD / PixAI taggers)

Tagger models add extra options: show probabilities, a minimum probability threshold, a maximum number of tags, and Tag filters — comma-separated rules to exclude or rewrite generated tags. Each comma-separated item is one rule.

• solo — remove the exact tag solo
• 1girl:person — replace the exact tag 1girl with person
• red* — remove tags starting with red (on word boundaries)
• *hair:wig — replace a trailing hair phrase with wig
• *hair*:wig — replace hair anywhere inside a tag

Exclude rules

Rule	Effect	Example matches	Example non-matches
phrase	Remove only the exact tag	hair	black hair, facial-hair
phrase*	Remove tags that start with the phrase	red, red eyes, red+vehicle	orange red, redirect
*phrase	Remove tags that end with the phrase	hair, black hair, facial-hair	hair style, chair
*phrase*	Remove tags containing the phrase	hop scotch, my hop, my-hop-tag	hope

Replace rules

Rule	Effect	Example output
source:target	Replace the exact tag	hair → wig
source*:target	Replace the matching prefix phrase	red eyes → blue eyes
*source:target	Replace the matching suffix phrase	black hair → black wig
*source*:target	Replace the matching internal phrase	big hair style → big wig style

Word boundaries are any non-alphanumeric character (spaces, +, -, etc.). Exact rules run before wildcard rules, and replace rules run before exclude rules. To use a literal comma or colon, wrap that side in double quotes (e.g. smile:":)"). If a replacement contains commas, it is inserted as multiple tags.

Advanced syntax reference

Both the image filter and the tag filters support prefixes, wildcards, and logical operators. The full details live in the collapsible sections below.

Image filtering syntax

Click a tag in the All Tags pane to filter for it, or type a filter in the Filter Images box for more control.

Full image filter syntax (click to expand)

Text prefixes

• tag: — images that have the term as a tag (tag:cat)
• caption: — the term appears anywhere in the caption string
• nl: — natural language content or presence (nl:true, nl:false, nl:"red coat")
• complete: — completion state (complete:true, complete:false)
• name: — the term appears in the file name
• path: — the term appears in the full file path
• No prefix — matches the caption or file path

Numeric prefixes (operators =/==, !=, <, >, <=, >=)

• tags: — number of tags (tags:=13, tags:!=7)
• chars: — characters in the caption (chars:<100, chars:>=30)
• tokens: — tokens in the caption (tokens:>75, tokens:<=50)

Spaces and quotes — wrap terms containing spaces in quotes: tag:"orange cat". Escape inner quotes with \, or mix quote types: tag:'orange "cat"'.

Wildcards — * matches any number of characters, ? matches one. tag:*cat matches orange cat, large cat, and cat.

Combining filters — NOT, AND, OR (lowercase works too). Precedence is NOT > AND > OR; use parentheses to change it, e.g. tag:cat AND (tag:orange OR tag:white).

Tag filtering syntax

The Search Tags box in the All Tags pane and the Filter Tag Library box in the Tag Library use the same style of syntax, applied to each tag.

Full tag filter syntax (click to expand)

Text prefixes

• tag: — tags whose name matches the term (wildcards allowed): tag:*hair matches long hair, short hair
• category: — tags in a category whose name contains the term (category:Character); use category:none or category:uncategorized for tags with no category
• No prefix — matches tags whose name contains the term

Numeric prefixes (operators =/==, !=, <, >, <=, >=)

• count: — how many times a tag is used (All Tags pane only): count:>10
• length: — number of characters in the tag name: length:<=3

Spaces/quotes, wildcards, and NOT / AND / OR work the same as in the image filter. Example: category:"Character" AND NOT tag:*_hair.

Marking images as complete

As you tag, you can mark images as complete to track your progress. This marker is stored separately by TagGUI Plus (in completion_cache.json) and is never written to your caption files, so it never becomes part of your dataset. It persists across sessions and folder changes.

• Mark/unmark with Ctrl+K / Ctrl+Shift+K, or via the right-click menu.
• Completed images show a green check in the thumbnail corner (toggle it in Settings).
• Press Ctrl+Shift+J to jump to the first incomplete image, or filter with complete:false.

Controls and context menus

Keyboard shortcuts are fully customizable in Settings → Keyboard shortcuts; the defaults are listed below.

Global

Action	Shortcut
Load / Reload Directory	Ctrl+Alt+L / F5
Settings	Ctrl+Alt+S
Exit	Ctrl+W
Undo / Redo	Ctrl+Z / Ctrl+Y
Find and Replace	Ctrl+R
Batch Reorder Tags	Ctrl+B
Tag Library	Ctrl+L
Danbooru / Gelbooru Wiki	Ctrl+D / Ctrl+G
Previous / next image	Ctrl+Up / Ctrl+Down
Jump to first untagged / incomplete image	Ctrl+J / Ctrl+Shift+J

Focus shortcuts

Focus	Shortcut
Filter Images box	Alt+F or Ctrl+F
Add Tag box	Alt+A
Image Tags list	Alt+I
Search Tags box	Alt+S
Start Auto-Captioning button	Alt+C

Images pane

• Zoom preview: mouse wheel or Ctrl++ / Ctrl+-; pan: drag or Alt+arrows; reset: double-click or Ctrl+0
• First / last image: Home / End
• Select multiple: Ctrl/Shift+click; Select all: Ctrl+A; Invert: Ctrl+I
• Switch between selected images without changing the selection: Left / Right
• Mark complete / incomplete: Ctrl+K / Ctrl+Shift+K
• Right-click menu: copy tags/caption/name/path, paste tags, open image or caption file, open in your configured editor, rename the image file, move or copy images to another folder, delete images, and mark complete/incomplete.

Image Tags pane

• Add a tag: type in Add Tag and press Enter; accept first suggestion: Ctrl+Enter
• Add a tag to multiple images: select them first, then add the tag
• Rename a tag: double-click or F2; delete: Delete; reorder: drag and drop
• Right-click a tag: copy it, view it on the Danbooru/Gelbooru wiki, or assign/clear its category.

All Tags pane

• Select a tag to filter images for it, or click to add it to selected images (depending on the click-action setting)
• Rename every instance: double-click or F2; delete every instance: Delete
• Right-click a tag: copy, wiki lookup, assign/clear category.

---

Original app by jhc13. This fork's additional changes were vibe coded with GitHub Copilot and are shared as-is.