Studio
An app node: create, edit, a gallery and a model manager in one window over ComfyUI. Each tool is one workflow file, so adding a model means adding a file.
I write custom nodes, turn workflows into pipelines and build full apps inside ComfyUI, tuned to run fast on the GPU you already have.
Five kinds of work, often in one project. You keep every file, with a guide for the next person who opens it.
The node your graph is missing: new controls, loaders for a new model, batch tools, a bridge to your own API or storage.
A whole app inside ComfyUI with its own interface, like Studio below. Your team works in forms and a gallery; the workflow runs underneath.
Clean graphs that someone else can read, and multi-step pipelines from brief to final file: stills to video, upscale, batch from a spreadsheet.
The same result in less time and less memory: quantized weights, step-reduced LoRAs, attention backends and offload settings matched to your GPU.
Broken installs, custom nodes that fight each other, missing models, an update that broke a graph the day before a deadline.
An app node: create, edit, a gallery and a model manager in one window over ComfyUI. Each tool is one workflow file, so adding a model means adding a file.

A node that turns a shot list into prompts for video models: shot size, camera move, lens and duration per shot, one combined prompt or one per shot.
One reference image, twenty scenes, and a face-match score for every frame, so consistency is a number and not a claim.
Every change is timed on your machine before and after. If it does not save time or memory, it does not stay.
The .json file, a screenshot of the error if there is one, your GPU or cloud setup, and what done looks like.
What is slow, what breaks and what is missing, in plain words, with a plan and a price before any work starts.
On your setup or a matching one. Speed changes come with timings, nodes come with tests.
Files, settings, a written guide and a recorded walkthrough, so your team can run it without me.
NVIDIA cards from 12 GB of VRAM up, from the RTX 30 series to Blackwell workstation cards, and cloud GPUs such as RunPod. Interfaces and nodes can be built and tested on a Mac.
Nodes ship with tests and a list of the versions they were checked on. When ComfyUI changes something they rely on, the tests show it before your team does.
Yes. You get the full source and the files with full usage rights for your project. Whether anything is open-sourced is your decision.
Yes, remotely on your server or workstation, or I build on matching hardware and hand over files that run on yours.
Price is agreed per project, after I have read the workflow and you know exactly what you get.
The .json file and one line on what it should do. You get a plain answer on what it needs and what it costs.