PixlStash 1.13

Draw what to change, and press Run

PixlStash 1.13 is about getting a workflow to run. The picture viewer's Edit tab has paint tools, so you can mark a part of a picture and have an inpaint workflow change only that part. Every workflow template your ComfyUI provides is now easily downloadable and PixlStash gets the models for you A workflow you add is wired to take its pictures, models and LoRAs from PixlStash and send its results back.

  • Paint tools in the Edit tab: rectangle, ellipse, lasso, brush, eraser and colour wand, then inpaint just the selection
  • Three inpaint workflows come with PixlStash, Qwen Image 2.1, Qwen Image Edit 2511 and FLUX.2 Klein 9B, and your own inpaint workflows work too
  • A Templates tab: every ComfyUI template, what it downloads and which of those files you already have
  • Download a missing model or LoRA in one press, with progress and cancel in the Tasks tab
  • Pick where downloads land in Settings › Models, and drop a model file on a folder to put it there
  • Runs wait for GPU memory instead of running out of it, and PixlStash steps aside for ComfyUI
  • Compare any two pictures on a before/after slider, and crop without ComfyUI

Inpaint from the picture viewer

Open a picture, go to the Edit tab and mark where the change goes with a rectangle, ellipse, lasso, brush, eraser or colour wand. Drag on from the wand's click to widen or narrow what it picks, as in GIMP. Grow, feather or fill the holes in the selection, say what should change, and press Run edit. Only the marked part is redrawn, the result joins the original's stack, and a before/after slider shows what changed. You can keep browsing while it runs.

Three inpaint workflows come with PixlStash: Qwen Image 2.1, Qwen Image Edit 2511 and FLUX.2 Klein 9B. Each cuts the selection out with some context around it, edit it at full resolution and stitch it back into the untouched picture. If their models are not on your shelf yet, the Edit tab offers to download them.

Or bring your own inpaint workflow

Any inpaint workflow in your library shows up in the Edit tab's list, and the selection you painted is the mask it gets. Here the same selection runs through a MOSS-Image inpaint workflow instead. With several pictures selected in the grid, Edit… applies one edit to all of them.

One edit after another

Drag a rectangle over the gnome's wooden rod and write "the wooden rod has an axe head so that it is a proper axe". Only the rod changes; the picture around it, down to the grass, is the original's. Then circle another gnome's mouth and ask for blue lipstick. Each edit is a new picture in the original's stack and the before/after slider shows exactly what each one changed.

The Workflows screen's Templates tab: a grid of ComfyUI template cards marked Ready to run or with how many files are missing, and the inspector for FLUX.2 Klein 9B Image Edit listing its three models, two of them on the shelf and one with a download button

Every ComfyUI template, one press from running

The Workflows screen has a Templates tab beside your own workflows. It lists every template your ComfyUI serves, ComfyUI's own and those of your custom-node packs, sorted with the ones you can run right now first. Each card says how many of its files you are missing, and the inspector lists them, with how much there is to download. With ComfyUI not running the tab still lists ComfyUI's own templates, from a copy that comes with PixlStash.

Add to my workflows takes one or several at once. A template you add is wired for PixlStash the way Open in ComfyUI is: it loads pictures from your library, models and LoRAs from your shelf, and sends what it makes back into your library.

The Tasks tab during a model download: the GPU memory bar, and a running row for a 19 GB Qwen Image checkpoint with a progress bar and a cancel button

Download missing models, into your folders

When a workflow is missing a model and the ComfyUI template list knows where to get it, a download button sits beside its name: in the Workflow tab, in the Run window, in the Edit tab, on a template and in a workflow set. A LoRA a run would leave out gets the same button.

There is no browser download to move afterwards and no models subfolder to guess. The file goes into the Download folder you set in Settings › Models, is checked against the template's hash where it lists one, and joins your model shelf. The Tasks tab shows its progress and can cancel it. Drop a model file on a folder in the Models screen and it goes there instead.

The Tasks tab with a ComfyUI inpaint run under Waiting for the GPU, needing 16.1 GB while 4.4 GB of 31.8 GB is free, and PixlStash's own CPU, RAM and VRAM use below

Runs wait for the GPU

A run that does not fit in GPU memory waits in the Tasks tab under Waiting for the GPU, saying what it needs, what is free and who holds the rest, and starts by itself once there is room. Start it anyway or cancel it. PixlStash pauses its own GPU work and unloads its models when ComfyUI needs the room.

A memory bar per GPU shows what PixlStash, ComfyUI and other apps hold. PixlStash learns how much each workflow needs from its runs, and a workflow that ran out of memory runs alone from then on.

Crop and compare, no ComfyUI needed

Crop a picture from the Edit tab to a square on the face, to 3:4, to the nearest training bucket, or to a frame you drag. The crop is a new picture in the original's stack, and the original stays as it was.

Compare any two pictures: select two in the grid and choose Compare, or right-click a second picture in the viewer's filmstrip. One is laid over the other with a slider.

Workflow sets follow their checkpoint

A workflow set made on a checkpoint now holds every model the pictures made with it used, and a model new pictures start using joins it by itself, marked New. What you add or remove by hand stays that way, and Reset to automatic undoes it. New workflow set starts from ComfyUI's templates, and a set from a template suggests the rest of its files, with a download for the ones you do not have.

Also in 1.13

  • Favourites for models, workflow sets and workflows: press H, and they come first in every list you pick from
  • Drop workflow .json files on the Workflows screen to add them
  • Pick your own picture for a workflow card with Choose thumbnail…
  • A workflow's version menu can duplicate an older version or open it in ComfyUI
  • Pick which GPU PixlStash uses on a machine with more than one
  • A missing checkpoint's picker sorts replacements by how likely they are to work
  • LTX-2 workflows saved from the ComfyUI editor now run

Upgrading to 1.13

Opening a library with 1.13 updates its database, and workflow sets you already have pick up the models their checkpoint's pictures used, marked New. Update the ComfyUI-PixlStash nodes from Settings › ComfyUI as well. The inpaint workflows that come with PixlStash also need the ComfyUI-Inpaint-CropAndStitch nodes in your ComfyUI. Back up the library folder first by copying it somewhere safe while PixlStash is stopped. See the upgrade guide for the steps for your install type.

Get PixlStash 1.13

Download the desktop app for Windows, macOS, or Linux. Or keep using pip, Docker, or the source build, and read the upgrade guide first.

Download Upgrade guide
Previous release: PixlStash 1.12
PixlStash 1.12

Pick a person, type a prompt, press Run

PixlStash 1.12 turns PixlStash into a ComfyUI orchestrator by turning the ComfyUI workflows behind your pictures into a library you can run. Every picture you made already says which workflow made it, so PixlStash groups them into a handful of cards, each covered with the best pictures it made. You can add new workflows directly from ComfyUI. Right-click a person, choose Create with LoRA, and type what you want to see. Their LoRA is put in for you, and the new pictures land back on that person.

  • A Workflows screen: every workflow your pictures were made with, as a card, with the one-offs folded away
  • A Recipe tab on every picture: the workflow, prompt, models, LoRA strengths, settings and seed that made it
  • One Run window wherever you start, with every value filled in and editable
  • Pick two people in the Run window, describe them both, and get one picture with both of them in it
  • One click installs the PixlStash nodes into your ComfyUI and links the two apps, from Settings
  • Open a workflow in ComfyUI, press Ctrl+S, and PixlStash keeps it as a new version
  • A workflow with a missing model is fixed in place from your model shelf
  • Text in pictures is searchable, and Apple Silicon Macs tag on the GPU

Connect ComfyUI and create pictures using LoRAs

One click in Settings › ComfyUI installs the PixlStash nodes into your ComfyUI. Those nodes are how ComfyUI loads your pictures and models and sends new pictures back, so this is the step that makes the rest of 1.12 work. PixlStash finds a ComfyUI running on this computer, installs the nodes (replacing an older copy), restarts ComfyUI through ComfyUI-Manager, gives the nodes a key of their own and checks that each app can reach the other. There is no token to copy, and ComfyUI's models folder joins your model shelf by itself. A ComfyUI on another computer gets the nodes from ComfyUI-Manager instead.

Then right-click a person or a picture set and choose Create with LoRA…. The Run window opens on the workflows for that LoRA's base model, with the LoRA already added and its trigger word offered for the prompt, and the new pictures are filed back under that person or set.

Two people in one picture

Press Run… on a workflow and pick two people under Person. Each brings their LoRA as a row of its own, with its strength and trigger word, so describe them both in the prompt and press Run. What comes back is one picture with both of them in it.

The Workflows screen: a grid of workflow cards, each covered with the pictures it made and labelled with its checkpoint, LoRAs and type, beside the inspector for the selected workflow

Your workflows as visual cards

The Workflows screen lists every workflow your pictures were made with, plus the files you add. The same graph run with another checkpoint or character LoRA stays one workflow, so thousands of pictures come down to a few dozen cards, and the ones you used once are folded away until you ask for them. Each card shows the pictures it made, its models and what kind of picture it makes. Sort by your ratings, by what you used last or by picture count.

From a workflow to its pictures and back

Show all pictures using this workflow opens your library on just those pictures, as a filter chip you can take off again. Going the other way, the picture viewer has a Recipe tab for anything made in ComfyUI, A1111 or Forge: the workflow it came from, the prompt, each model with the strength its LoRA ran at, and the seed. Videos made in ComfyUI get one too.

The Run workflow window with the prompt, a choice of person, LoRAs, size, steps, CFG, count, seed and checkpoint

A simple Run window

Press Run… on a workflow, Make more like this on a picture or Run a workflow on these on a selection, and the same window opens with every value filled in: prompt, people, LoRAs, size, steps, CFG, seed and checkpoint, and the rest under one disclosure. A field you change shows the value it started at, so you can put it back. Person takes several people, each with their LoRA as a row of its own, and any trigger word the prompt is missing is one click away. A run that cannot start says why, and names a missing model along with the folder it belongs in.

Keep track of generation and have them back in PixlStash

A run keeps its row in the Tasks tab with its progress and an abort, when you switch views or reload. When ComfyUI is done the pictures come into your library tagged and filed, and the workflow's card picks them up. A run that fails says which node failed and why.

A workflow opened in ComfyUI from PixlStash, with PixlStash loader nodes in the graph and a notice that it was saved to PixlStash as version 6 The ComfyUI menu with a PixlStash submenu offering Save to PixlStash and Save as new workflow in PixlStash

Open in ComfyUI and save back with Ctrl+S

Open in ComfyUI opens a workflow with PixlStash's loaders and savers in it, so its models load straight off your model shelf and what you queue comes back into your library. Press Ctrl+S in ComfyUI and the workflow gets a new version in PixlStash, with no file to name. The earlier versions are kept, and the version menu takes you back to any of them. Save as new workflow in PixlStash leaves the original as it was.

Pin parameters or make new ones available

Lock a parameter in the Workflow tab and the Run window stops asking for it. The + beside Parameters exposes any setting of any node in the graph under a name you choose, and a setting ComfyUI offers a list for, such as the sampler, is picked from that list.

The Open in ComfyUI button beside Run… in the Workflow inspector The Workflow inspector's ⋯ menu offering Open original in ComfyUI

Two ways to open a workflow in ComfyUI

The Workflow inspector has two ways in. Open in ComfyUI, the button beside Run…, opens the workflow adapted for PixlStash. Its nodes talk to PixlStash: Save Image and Save Video become the PixlStash savers, so what you queue comes back into your library; loaders whose file is on your model shelf load it from there; a Load Image holding a picture from your library becomes the PixlStash Picture Loader on that picture; and a PixlStash loader is added for trying the workflow on your people. Ctrl+S saves it back as a new version.

Open original in ComfyUI, in the inspector's ⋯ menu, opens the workflow exactly as it was made in ComfyUI, with its own nodes. It warns you when some of its models are only on the PixlStash shelf and not in ComfyUI's model folders, and names them. Both keep the layout you arranged in ComfyUI.

Repair workflows

A workflow whose checkpoint, VAE or text encoder has gone missing says so on its card. Pick another model of that kind from your shelf and the workflow keeps its name, pictures and settings. A missing LoRA, or a seed or text node from a pack your ComfyUI does not have, no longer stops a run: PixlStash says what it is leaving out and runs without it.

Workflow sets on the model shelf: a card per checkpoint, one opened to show its text encoders, VAE, LoRAs and upscalers

Workflow sets: define what fits together

The model shelf opens on a card for each checkpoint your pictures were made with. Open one and it lists every VAE, text encoder and LoRA that has run beside it, with how many pictures back each, so you can see what to pair with what instead of guessing. Make sets of your own, fill them from your pictures, and use a set to swap a workflow's models in one go.

Recipes: keep the look, not the graph

A recipe is a workflow plus the look: the prompt, the LoRAs with their strengths and any setting you changed. A workflow's Recipes tab lists every recipe your pictures were made with, with how many pictures each accounts for. Save as recipe keeps one, with the seed left out unless you tick it, and Run… starts from it. Export a workflow to share it and PixlStash takes out your prompt, seed and character LoRAs, and tells you what it left out.

Also in 1.12

  • Text in documents, receipts and screenshots is read in the background, and every search matches it
  • Apple Silicon Macs run tagging, search and face matching on the GPU, with the built-in tagger roughly twenty times faster
  • An Edit tab in the picture viewer runs an image-to-image or inpaint workflow without leaving it
  • pixlstash-mcp lets an AI agent search your library and read recipes, with a token from Connect AI agent
  • Move pictures and videos to another library
  • Models downloaded to your Downloads folder can be moved onto the model shelf by themselves
  • Suggest more pictures for a picture set, as you already could for a person

Why 1.12, and not 2.0

Workflows and recipes are the biggest change PixlStash has had, and only two things keep this from being 2.0. They are new, and they will get better as more people run them on their own libraries, so treat them as a preview and tell us what breaks. And one piece is still to come: 2.0 trains LoRAs with ai-toolkit from the people and sets in your library, so the whole loop from your pictures to a LoRA to new pictures of that person happens inside PixlStash.

Upgrading to 1.12

Opening a library with 1.12 updates its database, and the first start carries your workflows' names, notes, settings and saved recipes over to the new Workflows screen. Pictures and videos already in your library are read once more in the background, so their recipes show. Settings › Workflows is gone: add a workflow by dropping its file on the Workflows screen or into the watched workflows folder. Several features use the ComfyUI-PixlStash nodes that come with this version, and Settings › ComfyUI installs them for you. Back up the library folder first by copying it somewhere safe while PixlStash is stopped. See the upgrade guide for the steps for your install type.

Get PixlStash 1.12

Download the desktop app for Windows, macOS, or Linux. Or keep using pip, Docker, or the source build, and read the upgrade guide first.

Download Upgrade guide
Previous release: PixlStash 1.11
PixlStash 1.11

Point it at the folders you already have

PixlStash 1.11 is about getting started quickly with the pictures you already have, organised the way they already are. Load up a folder and it reads it as is. Your projects, people and sets are created from the existing folder names and nothing is moved, renamed or copied. Adding a library now happens in Settings instead of the terminal.

  • Your folders are read and PixlStash suggests from its content whether it is a project, picture set or a gallery of a person. You make the final decision.
  • Every library row has a menu to open, rename or stop using it. The latter does not delete anything.
  • A library can be given a folder layout, and new pictures are written where the layout says
  • A picture is only moved when its folder no longer describes it, and Move to match is offered, not done for you
  • Export straight into a folder on the machine running PixlStash, with no zip in between
  • A first pass over a big folder is noticeably quicker
The Add a library wizard, reading a folder of pictures and proposing what each folder level means PixlStash reading a folder of pictures, with each folder level labelled as a project, a person or a set
The Add a library dialog with a folder picked, PixlStash reporting what that folder is and what adding it would mean One dialog for every case
The folder mapping screen with three levels, each assigned to a project, person or set, and per-row overrides Your folders already say what they are
The review screen listing the projects, sets and people the folders will become, above a summary of what pressing the button does Nothing is written until you say so
Settings with several libraries listed and one row's menu open, offering switch, rename and stop using Rename it, switch to it, or stop using it
The first screen of a new library, offering a folder you already have, dropping pictures in, or connecting ComfyUI Three ways to fill a new library
The export panel offering Export to Folder alongside the zip download Export straight into a folder

Getting a library in, screen by screen

Import libraries with automatic assignment and a review screen. Rename, switch or remove libraries from the same place.

Upgrading to 1.11

Opening a library with 1.11 adds columns to its database. Back up the library folder first by copying it somewhere safe while PixlStash is stopped, as with every upgrade. API tokens and share links are left alone. See the upgrade guide for the steps for your install type.

Get PixlStash 1.11

Download the desktop app for Windows, macOS, or Linux. Or keep using pip, Docker, or the source build, and read the upgrade guide first.

Download Upgrade guide
Previous release: PixlStash 1.10
PixlStash 1.10

One PixlStash, several libraries

PixlStash 1.10 can hold more than one picture library and keeps one open at a time. Pictures, tags, scores and snapshots stay with their library, while your account and preferences stay with the installation. The release also adds a model shelf for the checkpoints and LoRAs on your disks, and a command line so backups and plugin installs can run from a script.

  • Several picture libraries in one installation, switched from Settings → Libraries. Each keeps its own pictures, tags, scores and snapshots
  • API tokens and share links belong to the library that created them, and Settings warns you before you switch away from live share links
  • A model shelf for the checkpoints and LoRAs in the folders you register, grouped by kind, base model or folder
  • Training runs from ai-toolkit are described where they lie, then imported with their sample images
  • A pixlstash-cli command line for libraries, backups, restores and plugins, so a backup can run from cron
  • The desktop app checks your library folder's permissions at start-up and offers to tighten them
PixlStash Settings with two registered picture libraries and the active one marked

Switching, and what stays behind

A library is a folder holding your pictures and their database. PixlStash keeps one open at a time and switches from Settings → Libraries. Your owner account, your preferences and the model shelf belong to the installation and follow you between libraries. API tokens and share links do not: they are pinned to the library that created them, stop working while another library is open, and work again when you switch back. Create a token in each library your scripts use, and expect Settings to tell you before you switch away from a library with live share links.

The PixlStash model shelf listing checkpoints and LoRAs grouped by kind, with size and date columns

A shelf for the models you already have

Register the folders where you keep checkpoints and LoRAs and PixlStash catalogues what is in them: the kind of file, the base model it was trained against, its size, and every folder a copy sits in. Group the shelf by kind, base model, folder or feature, sort it, and narrow it down to what you are looking for. Files are read where they are rather than copied into the library. A drive that is not plugged in says so instead of quietly showing nothing.

The PixlStash Training runs view with three discovered ai-toolkit runs selected for import

Training runs, described where they lie

PixlStash reads an ai-toolkit training run from the folder it was written to and describes it without touching a file: the checkpoints it produced, the base model, the rank, and the step each one stopped at. Import the runs you want and their sample images come with them. Versions of one subject group behind a single row, so a character you have trained four times is one entry with its history under it.

pixlstash-cli --help in a terminal, listing the libraries and plugins command groups and the exit codes

One command line for the installation

pixlstash-cli covers the jobs that point PixlStash at folders on your machine: adding and removing libraries, writing and reading back backups, and installing plugins. Every command documents itself, including the exit codes a script needs. Settings → Libraries shows the exact form for your installation with a copy button, and the desktop app can run it too, so PixlStash.AppImage cli libraries list works whether or not the app is open. Settings → Backend → Desktop → Shell command adds a pixlstash command in ~/.local/bin for any terminal.

The PixlStash model shelf's right-click menu with Move to Trash highlighted and a tooltip explaining Shift deletes permanently

Delete a model from the shelf

Remove a model from the selection pill, the row's right-click menu or the Delete key. Files go to your Trash, or are deleted for good with Shift held. Only your own model folders and PixlStash's own store are touched: a shared cache, or a drive that is not currently plugged in, is refused.

The PixlStash start-up screen listing library folders whose permissions are open to other accounts, with Quit and Fix it

A start-up check on folder permissions

If another account on the same computer can write to your library folder, that account could replace your pictures database or read your credentials. The desktop app notices at start-up, lists the folders it wants to change, and offers to fix them. Nothing is moved, renamed or deleted, and your own access is unchanged.

pixlstash-cli libraries --help listing list, create, attach, detach, relocate, backup, restore, prepare-legacy-identity and rename

Backups you can put in cron

libraries backup writes one library and its hub to a single archive you can read yourself, your password and API tokens included. libraries restore reads it back into a new folder and makes it the library that opens. Restore needs PixlStash stopped and never overwrites: your current config and hub are moved into a dated pre-restore- folder, and it prints the launch command for both the restored library and the one you had.

pixlstash-cli plugins available listing the published captioning plugins and image filters with their licences

Find a plugin by name, and check the one you are writing

plugins available lists what the plugin repository publishes: name, title, summary, and where they are declared the author and licence, with a star against anything you already have. Add a word to search. Until now the only way to learn a plugin's name was to guess one wrong and read the error. plugins test loads a captioning plugin the way the server does and checks its parameter schema is one the settings screen can draw, so a typo no longer costs a server restart. It imports your file, so that code runs with your permissions: it is a development aid and not a security scanner.

Upgrading to 1.10

The library upgrade is not backwards compatible: once a library has been opened by 1.10, an older version cannot open it again. Back it up first by copying the library folder somewhere safe while PixlStash is stopped. See the upgrade guide for the steps for your install type.

Get PixlStash 1.10

Download the desktop app for Windows, macOS, or Linux. Or keep using pip, Docker, or the source build, and read the upgrade guide first.

Download Upgrade guide
Previous release: PixlStash 1.9
PixlStash 1.9

Fix your mistake with one CTRL-Z

PixlStash 1.9 brings full undo and redo across the app. Soft-deletion, tags, descriptions, assignments, moves, and duplicate decisions all join one shared history. Step back through one change or several, then redo them when you change your mind again. Work quickly without making every decision final.

  • Full undo and redo, including a history menu for stepping back through more than one change. It is database backed and survives server restarts
  • A dedicated Duplicates view for reviewing, comparing, stacking, or keeping likely copies separate
  • Stack filters and Keep cover only for moving unwanted copies to the recoverable Scrapheap
  • Characters and picture sets can belong to several projects at the same time
  • Suggest more pictures of a person, then assign all results or only the ones you select
  • A new Privacy pane with two independent, off-by-default options: update checks and an anonymous install ID
PixlStash history showing several recent library changes with Undo and Redo controls

Undo reaches whole workflows

Use the toolbar, Ctrl+Z and Ctrl+Y on Windows and Linux, or the standard shortcuts on macOS. Soft-deleting pictures, changing tags or descriptions, assigning and moving images, and duplicate decisions all join the same history. Open the history menu when you want to inspect recent work or undo several steps at once. The safety net reaches whole workflows, not just simple field edits, including stacking duplicates or keeping them separate.

A queue for duplicate decisions

Duplicates has a permanent place in the sidebar and scans the view you are working in. Each group keeps its candidates together, explains why they match, and lets you compare them full-screen before you stack them or keep them separate. Exact copies are the starting point; optional near-duplicate tiers and a similarity threshold let you decide how far the scan should reach. Decided groups stay available, and Mixed stacks flags existing stacks whose pictures no longer hang together at the chosen threshold.

PixlStash Duplicates view with likely copies grouped for review
PixlStash picture menu with Keep cover only available for a selected stack Keep cover only preview showing three pictures moving to the recoverable Scrapheap in one undoable step

Clean up stacks without deleting originals

Filter any grid to stacked or unstacked pictures. When a stack is settled, Keep cover only previews exactly what will happen, keeps the cover, and moves every other picture in the stack to the Scrapheap. No original is deleted from disk by the action, and you can undo it or restore a picture from the Scrapheap.

One PixlStash character assigned to more than one project

One person or set, several projects

Characters and picture sets no longer have to be duplicated or moved back and forth when they belong in more than one project. Add the same item to several projects, see its pictures in every one, and remove one membership without disturbing the others. Existing project assignments carry over unchanged.

Suggest more pictures of a person

Open a person's context menu and ask PixlStash to rank the library by face likeness. Review the result in the grid, adjust the match if needed, then assign every result or just your current selection to that person in one action.

PixlStash Privacy settings with independent controls for update checks and anonymous install telemetry

Optional, limited telemetry

The new Privacy pane can send a random anonymous install ID once a day, together with your install type and whether this was a new install. It never includes your images, tags, captions, filenames, searches, or file paths, and the ID can be replaced whenever you like. It is off by default, stays off when you upgrade, and can be turned off again at any time. If you are comfortable enabling it, we hope you will: it gives us a much clearer picture of whether PixlStash remains useful after the download. The privacy policy shows the exact message and retention details.

A clear choice when you upgrade

Existing users who already enabled update checks see a one-time explanation of the anonymous install ID. The dialog shows the complete request, explains every field, and makes declining as straightforward as opting in. Nothing about the existing update check changes either way, and the choice remains available in Settings → Privacy.

PixlStash upgrade dialog explaining the optional anonymous install ID and offering No thanks or Add the random number
PixlStash Mixed stacks view showing stacks flagged for review

Review stacks that no longer fit

Mixed stacks finds existing stacks whose pictures fall below your chosen similarity threshold. Compare the members, keep the stack when it is intentional, or separate pictures that do not belong.

PixlStash grid filtered to show stacked pictures

Show stacked or unstacked pictures

Filter any grid by stack state to focus on pictures that have already been grouped or find the ones that still need attention.

Upgrading from 1.8.0 or earlier

PixlStash clears every API token as part of the security change released in 1.8.1. Create replacement tokens in Settings, recreate any share links, and enter your public URL and ComfyUI URL again. If you already upgraded to 1.8.1 or later, this has already happened and your current tokens are left alone.

Get PixlStash 1.9

Download the desktop app for Windows, macOS, or Linux. Or keep using pip, Docker, or the source build.

Download Upgrade guide
Previous release: PixlStash 1.8
PixlStash 1.8

A justified grid layout

PixlStash 1.8 adds a justified grid. Thumbnails keep their own aspect ratio and each row is filled to the full width of the grid, instead of every picture being cropped to a square. A single thumbnail-size control, from Tiny to Huge, replaces the columns slider and sets the size in both layouts. The square grid is still available.

  • A justified thumbnail layout, and one Tiny to Huge size control in place of the columns slider
  • Imports stream to the server and finish in the background, so closing the tab does not cancel them
  • Delete forever in the Scrapheap now removes the files from disk, behind a preview and a type-to-confirm dialog
  • An Agreement chart in the stats panel plots your star ratings against the Smart Score
  • Every API route goes through one deny-by-default authorization gate
  • Signed Windows installers since 1.8.2
The PixlStash justified grid, with thumbnails at their own aspect ratios filling each row to the full grid width

The justified grid while scrolling

Rows are filled to the grid width at every scroll position, and the date pill shows the date range of the pictures currently on screen. Selection, rating, and keyboard navigation work the same as in the square grid.

The PixlStash Appearance settings with a Square and Justified thumbnail layout picker

Choosing the layout

The layout is set in Settings under Appearance. The same pane sets the sidebar thumbnail size and switches the sidebar between full width and a narrow icon dock. The grid thumbnail size is in the toolbar, next to the sort and filter controls.

Scrapheap delete and auto-empty

Delete forever in the Scrapheap now removes the files from disk. It lists what will be deleted and asks you to type to confirm. Auto-empty is off by default: nothing is removed on a timer until you set a retention window of 30, 60, 90, or 120 days in Settings. Installs upgrading from 1.7 stay off until you set one. Originals in reference folders are never auto-purged.

Background imports

Imports stream to the server and complete there. Closing the tab or navigating away does not cancel them. A running import appears in the task manager with progress, and can be aborted from there.

Agreement chart

The stats panel has a new chart that plots your star ratings against the Smart Score. Cells off the diagonal are pictures where the two differ: rated high and scored low, or the reverse. Clicking a cell opens those pictures in the grid.

Smart Scores are recomputed on upgrade

The whole library is re-scored in the background the first time you open 1.8, so the grid may re-order compared to 1.7. Two things changed. The built-in reference points used to score a library with few ratings of its own had stopped loading and now load again, so if you have rated only a handful of pictures your scores will change more than most. The scoring weights are also rebalanced so the top of the 1 to 5 range is reachable. The recompute runs in small batches and does not re-run any AI models. Originals and snapshots are not modified.

One authorization gate for the API

Every API route now goes through a single deny-by-default authorization gate that enforces object access from a declared policy. A route with no declared policy is refused rather than served, so a share link or API token can only reach what it was granted. This release also fixes snapshot restore hard-deleting files that were added after the snapshot was taken, and stops deletions removing pictures in locked sets.

Get PixlStash 1.8

Download the desktop app for Windows, macOS, or Linux. Or keep using pip, Docker, or the source build.

Download Upgrade guide
Previous release: PixlStash 1.7
PixlStash 1.7

Fix your tags, one clear decision at a time

PixlStash 1.7 adds a tag review queue. It finds the tags most likely to be wrong or missing in your library and walks you through them picture by picture. Each card shows why the picture was flagged and asks one question: should this picture have the tag or not. Answer with a click or a single key, and the tagger learns from your decisions.

  • A ranked queue of the tags most worth reviewing, with estimated wrong and missing counts per tag
  • Lockable picture sets: freeze a set as a read-only eval or training set
  • Object detection that draws labelled boxes and exports them as COCO or Ideogram JSON
  • A smarter, calibrated Smart Score with new checks for compression artifacts, noise, and watermarks
  • A refreshed visual design across the whole app
PixlStash 1.7 tag review card asking whether a picture should have a tag, with similar images shown for context

See which tags need attention

The tag health view ranks every tag in your library by how much a review is likely to fix. It estimates how many pictures have the tag wrongly, how many are missing it, and where the tagger disagrees with your past decisions. Start a review straight from the row, and filter by project, set, or character to keep the session focused.

Tag health table ranking tags by review priority with estimated wrong and missing counts
Sidebar context menu locking a picture set in PixlStash

Lock a set when it is done

Lock a picture set to freeze its pictures' tags, descriptions, and scores. Locked sets are read-only until you unlock them, and a lock badge shows wherever the pictures appear. Useful for eval sets and training data that must not drift after you have curated them.

Object detection in PixlStash drawing labelled bounding boxes around a camera, a car, and starfish

Find the objects in your pictures

A new Segment action detects objects and draws labelled boxes in the grid and the picture view. Export the boxes as COCO or Ideogram JSON to use them in your own training and annotation pipelines.

A cleaner look across the app

The whole app has been restyled on a design token system, with consistent type, colour, spacing, and icons. Dialogs, menus, and settings follow one visual language, in both light and dark themes.

The refreshed PixlStash settings dialog in the new visual design

A Smart Score you can trust more

Smart Score is now calibrated and precision aware. It weighs each defect tag by how confident the tagger is, groups related defects so correlated flaws do not count twice, and adds three new full-image checks for compression blockiness, noise, and watermarks. Your whole library is re-scored in the background the first time you open 1.7, so the grid may re-order compared to 1.6. That is expected. Your originals and snapshots are untouched.

Previous release: PixlStash 1.6
PixlStash 1.6

PixlStash comes to the desktop

PixlStash 1.6 introduces the PixlStash Desktop app for Windows, macOS, and Linux: one download that wraps the local server in a native window. It runs offline out of the box and opens straight into your library. No terminal, no Docker, no login screen.

Already running PixlStash from pip, Docker, or the Windows installer? The desktop app picks up your existing library, so nothing moves and nothing is lost.

  • A native app for Windows, macOS (Apple Silicon), and Linux: one download, simple install.
  • Runs offline out of the box and opens straight into your library, no login
  • GPU acceleration added on demand. NVIDIA CUDA and experimental AMD ROCm
  • Lives in the system tray and keeps serving in the background when you close the window
  • Optional remote access so other devices on your network can open your library
  • Don't worry: PixlStash server isn't going anywhere. It remains and will remain a clear focus.
PixlStash 1.6: the desktop app showing the library grid in a native window

Match the speed to your hardware

The app ships with a built-in runtime that works on any machine: CPU on Windows and Linux, and Apple Silicon GPU acceleration (Metal) on macOS out of the box. On Windows and Linux, if you have a discrete GPU, add acceleration on demand from the Backend settings and choose from NVIDIA CUDA or experimental AMD ROCm libraries download straight from PyPI into a private overlay, and the bundled runtime is never touched. The same panel turns on optional remote access so other devices on your network can open the library in a browser in the same way as PixlStash server.

PixlStash desktop Backend settings: compute acceleration and remote access

Runs offline, lives in your tray

Everything is local. The bundled runtime means the app works with no internet connection, and it opens directly into your library with no login prompt because it talks to its own private local server. Close the window and PixlStash keeps running in the system tray, so your library, and any remote access you enabled, stays a click away. The pip, Docker, and source installs are all still fully supported; the desktop app is simply the easiest way to run PixlStash on your own machine.

Previous release: PixlStash 1.5
PixlStash 1.5

Snapshots and restore with a full API

PixlStash 1.5 puts a safety net under all your organising work. Automatic, compressed snapshots capture your library's metadata: tags, scores, faces, descriptions, and how everything is organised, so any mistake is a few clicks from undone. Roll all your metadata back, or restore just the pictures you select.

(Your image files are never touched.)

  • Automatic daily / weekly / monthly snapshots, compressed with zstd
  • Full metadata restore with an automatic safety snapshot
  • Selective restore of just the pictures you choose
  • Breadcrumb navigation and drag characters & sets between projects
  • A clean snapshot & restore REST API
PixlStash 1.5: snapshots panel with daily, weekly, and monthly restore points

Restore only what you choose

A full restore rolls all your metadata back to a point in time, but you don't always want that. Select any pictures and restore just their metadata from a snapshot, leaving everything else exactly as it is.

Selective restore previewing exactly which selected pictures will change
Breadcrumb navigation showing the path to the current view in PixlStash

Find your way with breadcrumbs

A breadcrumb trail above the grid always shows where you are: project, character, or set. This allows you to jump back up the hierarchy with a single click. No more guessing how you reached a view or navigating back from scratch.

Dragging a character onto a different project in the PixlStash sidebar

Drag characters & sets between projects

Reorganise on the fly: grab a character or picture set in the sidebar and drop it onto another project. Their pictures are re-assigned for you. The move is reference-aware, so a picture that still belongs elsewhere stays put. Reshaping large libraries is now a drag away.

Snapshots without the disk bill or the wait

A snapshot copies only your metadata database, never the image files, which would be wildly impractical to duplicate over and over. That database is compressed with zstd for roughly a 3× saving, yet still keeps the expensive parts, CLIP embeddings and face data, so a restore comes back fully populated with no re-tagging or re-embedding pass afterwards. Retention follows a true grandfather-father-son policy: the most recent 7 daily, 4 weekly, and 12 monthly snapshots are kept and the oldest are pruned for you.

The snapshot & restore API

Everything the snapshots panel does is a plain REST call, so you can script metadata snapshots and restores straight into your own pipelines. Authenticate with an owner token and talk to the endpoints under /api/v1.

Every request carries an owner token: Authorization: Bearer <token>
GET /snapshots

List every restore point, newest first.

Request 
curl -H "Authorization: Bearer $TOKEN" \
     https://your-server/api/v1/snapshots
Response 
[
  {
    "id": 42,
    "kind": "DAILY",
    "label": null,
    "created_at": "2026-05-31T06:42:37Z",
    "picture_count": 1030,
    "byte_size": 4523754,
    "is_compatible": true
  }
]
POST /snapshots

Take a manual snapshot right now, optionally labelled.

Request 
curl -X POST https://your-server/api/v1/snapshots \
     -H "Authorization: Bearer $TOKEN" \
     -H "Content-Type: application/json" \
     -d '{"label": "before big cleanup"}'
Response 
{ "id": 51, "kind": "MANUAL", "label": "before big cleanup", "picture_count": 1030 }
GET /snapshots/{id}/restore/preview

Dry-run a full restore. It counts only what actually changes.

Response 
{
  "summary": {
    "pictures_to_revert": 12,
    "pictures_to_recreate": 3,
    "pictures_to_delete": 1,
    "pictures_unchanged": 1014,
    "missing_files": 0
  },
  "warnings": []
}
POST /snapshots/{id}/restore

Restore all metadata to this snapshot. A safety snapshot is taken first.

Request 
curl -X POST https://your-server/api/v1/snapshots/42/restore \
     -H "Authorization: Bearer $TOKEN"
Response 
{ "snapshot_id": 42, "resource_type": "full", "missing_files_count": 0, "errors": [] }
POST /snapshots/{id}/restore/batch

Selective restore. Bring back only the resources you list.

Request: curl 
curl -X POST https://your-server/api/v1/snapshots/42/restore/batch \
     -H "Authorization: Bearer $TOKEN" \
     -H "Content-Type: application/json" \
     -d '{"resources": [{"type": "picture", "id": 8123},
                        {"type": "picture", "id": 8124}]}'
Request: Python 
# pip install requests
import requests

r = requests.post(
    "https://your-server/api/v1/snapshots/42/restore/batch",
    headers={"Authorization": f"Bearer {token}"},
    json={"resources": [{"type": "picture", "id": 8123}]},
)
print(r.json())  # {'snapshot_id': 42, 'upserted_count': 1, 'errors': []}
Response 
{ "snapshot_id": 42, "upserted_count": 2, "errors": [] }
Browse the full interactive API reference →
Previous release: PixlStash 1.4
PixlStash 1.4

Search by picture, face, and meaning

PixlStash 1.4 adds powerful new ways to find pictures. Reverse image search is now a right-click away in the grid and with a full API, wired straight into your generation workflows with the new ComfyUI-PixlStash nodes.

PixlStash 1.4: context menu with reverse image search and find similar faces

Search from any ComfyUI workflow

The ComfyUI-PixlStash nodes expose PixlStash search inside ComfyUI. Drop in a reference image and the PixlStash Likeness Search node returns the closest matches from your library. Switch between face_search and picture_likeness modes to match on a person or on an overall picture. Store the pictures straight back into your library with the Picture Saver node.

ComfyUI graph using the PixlStash Likeness Search node in face search mode
ComfyUI graph using the PixlStash Semantic Search node with a text query

Semantic search by description

Describe what you're looking for in plain language: "man searching for photos" and the PixlStash Semantic Search node returns matching pictures ranked by relevance, with a similarity threshold and result limit you control.

ComfyUI graph batching multiple references into a single PixlStash likeness search

Combine several concepts

Batch multiple characters, sets, and individual pictures into a single likeness search and combine them with mean or min. Surface the pictures closest to all of your inputs at once, not just one. Old man drinking beer by a river? No problem!

Even faster grid loading

The grid keeps getting quicker. 1.4 builds on the 1.3 loading overhaul with further optimisations, so even very large libraries feel more responsive as you scroll, filter, and search.

Previous release: PixlStash 1.3
PixlStash 1.3

Snappy grid, smarter tagging

The image grid now loads visually very quickly and fills in details asynchronously. Even at 30 000+ images it feels snappy. Plus full JoyCaption support, selectable tag and description engines, and a URL that remembers where you were.

  • Dramatically faster grid: visual content appears immediately
  • JoyCaption support for tagging and image descriptions
  • Choose your tag or description engine per-request
  • Persistent view URLs: refresh and stay exactly where you were
  • Drag tags to accept or reject them in the tag panel
PixlStash 1.3: tagger and description engine settings
Tagger and description plugin settings in PixlStash 1.3

Selectable tagger & description engines

The settings page now lets you choose which plugin handles automatic tagging and which generates image descriptions. Each engine has its own parameter controls, so you can tweak precision, prompt style, and thresholds without leaving the settings panel.

Browser URL bar showing a persistent PixlStash view URL

Persistent view URLs

Every view, whether grid, character, picture set, or project, has its own URL. Bookmark it, share it with a teammate, or simply refresh and land exactly where you left off. No more navigating back from scratch after an accidental reload.

One-click tagging & descriptions

Right-click any image (or use the selection toolbar) to re-tag or regenerate a description using the engine of your choice. A hover submenu lists every available model so you can pick the right one for the job without touching the settings page.

Context menu with Tag automatically and Generate description submenus

Snappy grid loading

The image grid renders the visual content immediately and fills in metadata such as scores, tags, and faces asynchronously in the background. The result is a dramatically snappier feel: thumbnails appear before the first server response has even finished, and the grid stays responsive while data loads in. This is the single biggest day-to-day improvement in 1.3.

Previous release: PixlStash 1.2
PixlStash 1.2

A whole new look, and picture sharing

The biggest visual overhaul since launch: a redesigned sidebar, new toolbar, and context menus throughout. Plus, share pictures, sets, characters and projects with a single link.

  • Share pictures and collections with read-only links
  • Redesigned sidebar and toolbar
  • Context menus across the grid and sidebar
  • Faster background processing: tagging & face extraction
PixlStash 1.2: redesigned sidebar and toolbar

Picture Sharing

Share individual pictures, sets, characters, and projects with anyone using a read-only link. No account is required for viewers. Tokens are scoped, revocable, and never expose your full library. Ideal for sharing work-in-progress with clients or collaborators. Optionally add your own watermarks to shared pictures for branding.

Sharing dialog in PixlStash showing a generated share link
Shared picture view as seen by a recipient

Recipient view

Recipients see a clean, read-only view of exactly what you shared. No login, no noise, just the content you chose to share.

New PixlStash toolbar design

New Toolbar

The toolbar has been redesigned for clarity. Search, filters, and sort controls are better organised, and the new layout scales more naturally on smaller screens.

Context menu in the PixlStash image grid

Context Menus

Right-click anywhere in the grid or sidebar to access actions for the item under the cursor. Rate, tag, assign, move, share, or delete without leaving your current view.

Coloured icon picker for picture sets in PixlStash

Custom Icons for Sets

Give each picture set its own coloured icon instead of relying on the automatic thumbnail. Instantly recognise your sets at a glance, no matter how similar the contents look.

Redesigned PixlStash sidebar

Redesigned Sidebar

People, sets, projects, and folders are easier to navigate. The new sidebar is more compact, loads faster, and shows more at a glance.

Task manager in the statistics sidebar showing processing throughput

Faster Background Processing

Tagging and face extraction are significantly faster thanks to bulk database operations and smarter grid update scheduling. Watch real-time throughput in the statistics sidebar task manager.

Security Vulnerability Notifications

PixlStash now checks for known security issues in the ChangeLog and alerts you directly in the UI when a fix is available along with the security level and affected versions. Stay informed and keep your installation secure with ease.

Previous release: PixlStash 1.1
PixlStash 1.1

More control over your image folders

This release adds reference folder support, letting you keep your existing folder structure while still benefiting from PixlStash's powerful organisation and metadata features.

  • Reference folders and import folders managed within the UI
  • Context menus in image grid and main sidebar
  • Statistics sidebar for rapid overview
PixlStash screenshot with reference folders, context menu, and statistics sidebar
Tag distribution and confidence statistics sidebar in PixlStash

Tag Statistics

Get useful statistics about your picture sets. See tag distribution and tag prediction confidence. A powerful way to understand your training sets.

Picture statistics including image size and score distribution

Picture Statistics

See statistics about your pictures including size distribution and score distribution. Filter image views by clicking on elements in the sidebar. Quickly find outliers to review or delete.

Boolean Set Operations for People & Sets

Hold Ctrl and click names in the People or Sets sidebar to build a multi-selection, then choose how to combine them using the toolbar at the bottom of the grid. Union shows all pictures from any selected person or set. Overlap finds pictures where every selected person appears together. Difference subtracts one set from another, useful for finding pictures unique to a base set. Unique (XOR) returns pictures that belong to exactly one of the selections, letting you quickly spot what each person or set contributes exclusively.

Multi-select with Overlap mode showing pictures containing both selected people

Looking for full release details? Read all release notes on GitHub