ComfyUI and generative media pipelines
Generative image and video pipelines that run as dependable production systems, not as a fragile workflow on one person's machine.
The problem
A ComfyUI workflow that works on a creative’s machine is a prototype. Putting it into production means pinned model and node versions, queueing, retries, storage, and outputs that can be reproduced weeks later.
What gets built
- Workflow engineering: graphs restructured for reliability and speed, parameters exposed cleanly.
- Custom nodes in Python where the built-in ones fall short.
- ComfyUI as a backend: an API in front of a queue of GPU workers, with job tracking and storage.
- Batch production pipelines for catalogues, variants and video, with review steps.
- Product imagery that starts from your real product photos, so the product stays exactly as it is and only the setting changes.
What you get to know
Measured cost and speed per image or clip on the hardware you would run, whether outputs are reproducible from their recorded settings, and what happens when a job or a GPU fails. LAB/004 shows this for one pipeline: about 6 seconds per 1024 × 1024 image on one consumer GPU, identical output for identical input, and a killed server recovering every job.
When this is the wrong tool
If a few images a month are needed, a designer with a hosted tool is cheaper than any pipeline. If every image must show a specific product with exact details, text-to-image alone will not hold them; the pipeline has to start from real photographs, and a person still approves what is published. Some uses are restricted by model licences; that is checked before anything is built.
How an engagement starts
With a fixed-scope assessment of your current workflow or brief: the models and licences that fit, a measured sample run on representative inputs, and what production would cost per output.
Have a project like this?
- I read your message and reply personally, normally within two working days.
- Any questions are clarified by email. A short call only if required.
- A written proposal for a fixed-scope first step, with a fixed price.
Evidence and further reading
Generative Media Pipeline
An image-generation pipeline (ComfyUI and FLUX.1-schnell on one consumer GPU) tested like production software: same input, same image; a queue that finishes; and a crash that loses nothing.
How an engagement works
Every project starts small and fixed in scope, so you see evidence before you commit to more.
Feasibility assessment
A short, fixed-scope look at your problem, your data and the options.
A written assessment and a clear recommendation, including "don't build this" when that is the honest answer.
Prototype
A working system on your real data, not a demo on someone else's.
The prototype plus an evaluation report with measured results.
Build
The production system, integrated with what you already run.
Tested, documented code that you own, with a proper handover.
Support
Monitoring, evaluation reruns and improvements as your data changes.
Agreed scope and response times, no lock-in.