A pack of custom ComfyUI nodes that wire your workflows straight into a PixlStash vault. Load images by project, set, or character; search by face, picture, or meaning; gate generations by likeness; and save the results back with full metadata with no manual exporting in between.
Install the nodes and link the two apps from PixlStash in one click, then drive PixlStash from ComfyUI: pull matching faces out of your library, and extend a library image beyond its borders.
Settings › ComfyUI in PixlStash installs these nodes into your ComfyUI and links the two apps. It finds a ComfyUI running on this computer, installs the nodes, restarts ComfyUI through ComfyUI-Manager and gives the nodes a key of their own, so there is no token to copy. Then right-click a person, choose Create with LoRA, and the new pictures come back filed under them.
Open a workflow in ComfyUI from PixlStash and its loaders read your model shelf. Press Ctrl+S, or use the PixlStash menu, and the workflow is saved back to PixlStash as a new version.
Use the PixlStash ComfyUI nodes to search your library for pictures matching a face, then pull them straight into your workflow to edit them with no manual exporting in between.
A ready-made ComfyUI node and workflow lets you load an image, outpaint it, extending it beyond its original borders and saving the result back to PixlStash.
Linking ComfyUI from PixlStash's Settings › ComfyUI installs these nodes and sets up their connection for you. Otherwise, install them below and configure the connection (URL and API token) once in ComfyUI Settings > PixlStash. The nodes read it at runtime. They never appear as node widgets or in saved workflow JSON.
Selects a project from your vault and outputs a
PIXLSTASH_PROJECT wire that scopes the operations of
every node downstream.
Selects a set within a project, outputting
PIXLSTASH_PROJECT and PIXLSTASH_SET
wires. Reference-character sets are excluded from the dropdown.
Selects a character from your vault, outputting
PIXLSTASH_PROJECT and
PIXLSTASH_CHARACTER wires.
Requires PixlStash v1.2.1+.
Loads images as IMAGE and MASK tensors.
Picker mode opens a thumbnail browser to choose
exactly the images you want; browse mode fetches
automatically from any connected project, set, or character
filter. Context wires pass straight through to a downstream saver.
Uploads images and optionally assigns them to a project, set, and/or character. Supports embedded workflow metadata in PNG output and returns the IDs of successfully imported pictures.
ComfyUI's Save Video, saving to PixlStash: the same format, codec and encoding options, with the video optionally filed under a project, set or character. Can embed the workflow in the file. Requires PixlStash v1.2+.
Search by likeness to a face, or combine several reference images into a single embedding query (e.g. "an old man and a young man drinking beer"). Filter by project, character, and set by wiring those inputs. Requires PixlStash v1.4+.
Pull images out of the vault by meaning. Type a text string and the node uses PixlStash's semantic search to return the closest matches. Filter by project, character, and set. Requires PixlStash v1.4+.
Keep only the generations that actually match a reference
character. Wire in a Character Loader plus a batch of images, set
a likeness threshold, and the gate splits the batch into
accepted and rejected streams (with
counts), so you never waste compute polishing an off-model
render.
Requires PixlStash v1.4+ and a running face-extraction
worker.
The same split, judged on whole-picture likeness to a reference
set rather than a face. combine decides how the
set's members vote: min means match every member,
max means match any.
Requires PixlStash v1.4+.
Pick models from your PixlStash model shelf in a thumbnail grid instead of a dropdown of hundreds of filenames. Each loader has the same inputs and outputs as its ComfyUI built-in, so it drops straight into an existing graph. Files are used in place when ComfyUI and PixlStash share a disk, and fetched once and cached otherwise (checkpoints must be readable locally). The shelf needs an owner token.
Every loader opens the same grid, searchable by name, filename or trigger word. For LoRAs, People who fit lists the people with an adapter for the workflow's base model, by face, with a choice when more than one of theirs fits.
Applies a LoRA from the shelf and adds a
trigger_words output. Wire in a Set or Character
Loader to see only their adapters, and base_model
narrows the grid to one family: Krea 2 shows the Krea 2,
Raw and Turbo adapters alike.
Requires PixlStash v1.10+.
Several LoRAs in one node, one row each with its picture, its
person, its trigger word and two strengths, applied top to bottom
as chained loaders would be. Set show to
People who fit to pick people instead of files. Up to
eight rows.
Requires PixlStash v1.12+.
A checkpoint, a VAE, or one or two text encoders off the shelf,
each showing the model's name and icon on the node. The
Checkpoint and CLIP loaders also read .gguf files,
through a bundled copy of ComfyUI-GGUF.
Requires PixlStash v1.10+.
A ControlNet, or an upscaler such as 4x-UltraSharp
(.pth or .safetensors), off the shelf,
so a workflow no longer needs the file in ComfyUI's own folder.
Requires PixlStash v1.12+.
Loads one of your workflow sets in one node: its checkpoint, with the set's text encoders and VAE in place of the checkpoint's own. Wire it into an Adapter Loader and that loader's grid shows only the set's LoRAs. Requires PixlStash v1.12+.
Ready-to-load workflow JSON files ship in the examples directory. They use only PixlStash nodes plus core ComfyUI nodes, so they load on any install. Click a screenshot to enlarge it.
The starting point: load images from a project, set, or character and save the results straight back into your vault.
Open workflow JSON →End to end: find source images by face or meaning, run them through an edit pipeline, and save the edited results back to PixlStash.
Open workflow JSON →Load a picture from the vault and edit it with a plain-language instruction ("replace her top with a blue wooly jumper"), then save it tagged to the same character.
Open workflow JSON →Extend a loaded vault image beyond its original frame, then save the wider result back with its metadata intact.
Open workflow JSON →Upscale a vault image and write the higher-resolution result straight back to PixlStash with metadata preserved.
Open workflow JSON →Run it end to end: generate with a character LoRA, gate by face likeness, then upscale the accepted frames and save accepted and rejected streams to separate sets.
Open workflow JSON →Because the Picture Saver writes straight to PixlStash, everything you generate or edit shows up organised by project, set, and character, ready to search, tag, and curate.
Add the PixlStash nodes to your ComfyUI install, then point them at your library.
With ComfyUI running on the same computer, PixlStash 1.12 finds it, installs the nodes, restarts ComfyUI through ComfyUI Manager and links the two apps. No URL or token to enter.
Click Manager in the ComfyUI menu, then open the Custom Nodes Manager. Don't have it yet? Install ComfyUI Manager first.
Search for ComfyUI-PixlStash and click
Install on the matching result.
Restart ComfyUI when the install finishes, then set your PixlStash URL and API token under Settings > PixlStash.
Restart ComfyUI after cloning, then configure your PixlStash URL and API token under Settings > PixlStash.
| Setting | Description |
|---|---|
| URL | Base URL of your PixlStash instance. |
| API Token | A token with the required read or write scope. The search and loader nodes need read scope; the Picture Saver and Face Likeness Gate need write scope. |
| Verify SSL | Whether to validate the server certificate. |
Credentials live only in ComfyUI's settings. They are never written
into a saved workflow, so you can share a .json safely.
Need a token? Create a scoped one in PixlStash under user settings.
See the getting-started tour and the
API reference.