Browse categories
English
$
USD
Connect to your AI
I will deploy your comfyui workflow as a runpod serverless API with docs
Robin
About this gig
Your ComfyUI workflow works on your machine. Now your app needs to call it as an API. I deploy it to RunPod Serverless on your account and hand over everything you need to keep it running.
What you get:
- A Docker image with your custom nodes and pinned dependencies
- Models on a Network Volume, or baked into the image when cold start matters more
- An endpoint tested with real requests, not just a green build
- Image and video results as base64, or uploaded to your S3 bucket
- A written API doc with working Python and curl examples
- Notes on GPU choice, VRAM and every issue I hit
Experience: I have deployed multi-workflow image and video pipelines (Wan Animate, skin upscaling, FLUX) for overseas teams, tested on 4090, 5090 and B200, including large video outputs and private storage.
How it works (no calls needed):
1. Send the workflow JSON, sample inputs and model list
2. I audit nodes and models and confirm the scope in writing
3. You get the endpoint, docs and a test report
You pay RunPod directly for GPU and storage. I do not deploy NSFW or face-swap workflows.
Get to know Robin
Robin
AI video producer and ComfyUI deployment engineer
- FromChina
- Member sinceSep 2020
- Avg. response time1 hour
Languages
Chinese, English
I'm Robin. I make AI videos and turn ComfyUI workflows into APIs your app can call.
Video: property films with maps and 3D build sequences, AI-animated story episodes, motion ads with test variants, and animated explainers.
Deployment: ComfyUI image and video workflows on RunPod Serverless, with tested requests, API docs and handover notes.
Everything is in writing: a clear brief, sample frames or a test report, then timestamped feedback. No calls needed.
FAQ
Do I need a RunPod account?
Yes. The endpoint, image and volume stay in your account, so you own everything. You can invite me to your RunPod team, or I give you a ready image and a step-by-step guide to create the endpoint yourself.
My workflow has errors or missing models. Can you still deploy it?
I audit it first. Missing models and broken nodes are listed before I build, and repairs are quoted separately. Quality problems inside the workflow design are outside a deployment order.
Can you deploy video workflows?
Yes, including Wan 2.1 and 2.2 based workflows. Large video files are uploaded to your S3 bucket instead of being returned as base64.
How fast is the cold start?
It depends on model size, GPU and whether models sit in the image or on a volume. I test both options where it matters and write down the trade-off. I do not promise a fixed number before testing.
Which models do you support?
FLUX, Qwen-Image and Qwen-Image-Edit, Z-Image, SDXL, Wan, upscalers and most community custom nodes. You are responsible for having the right to use the models commercially.
Do we need a call?
No. Scope, progress and handover are all written. You get a test report and a runbook.
What happens after delivery?
Standard includes 3 days and Premium 7 days of bug fixes for the delivered scope. Changes to the workflow itself are a new order.
