GenAI Engineer for AI Image and Video Generation Pipeline (ComfyUI, Flux, SDXL, Turbo-Z, LoRA)

We are bringing AI image and video generation in house for our product and marketing work, and we need a GenAI engineer who lives inside ComfyUI to get us to a fully working, reusable environment. What we want at the end of this project: a clean ComfyUI installation on a GPU environment with Flux, SDXL and Turbo-Z checkpoints running, LoRA models loading correctly (plus guidance on training our own), a working video generation workflow (AnimateDiff, SVD or similar), generation exposed through API end points so our app can trigger it programmatically, and auto scaling GPU workers so the pipeline handles demand spikes and costs nothing when idle. Everything documented in a short guide so our team can generate images and videos on day one. Scope of work: ⦁ Install and configure ComfyUI on a GPU environment (RunPod, local GPU or cloud, we are open to your recommendation) ⦁ Add and verify Flux, SDXL and Turbo-Z model checkpoints with the right VAEs, samplers and settings, including a fast low step Turbo workflow for quick previews ⦁ Configure LoRA loading and provide a recommended workflow for training custom LoRAs on our brand assets ⦁ Install the essential custom nodes (ComfyUI Manager, ControlNet, upscalers, face fix and similar) ⦁ Build one text to image workflow (Flux, SDXL and Turbo-Z variants) and one video generation workflow ⦁ Expose generation through API end points (RunPod serverless or the ComfyUI API) so our application can trigger image and video generation programmatically ⦁ Configure auto scaling GPU workers: scale up on demand, scale to zero when idle, so we only pay for what we use ⦁ Export all workflows as JSON files and write a short usage guide with recommended prompts, settings and API examples What matters to us: You have done this before, end to end, and can show outputs. You know which checkpoints, custom nodes and settings actually work together instead of fighting CUDA errors for hours. You can explain your choices simply, because our team will run this after you hand it over. You should be comfortable with: ⦁ ComfyUI: installation, custom nodes, workflow design and troubleshooting ⦁ Flux, SDXL and Turbo class models (Turbo-Z, SDXL Turbo, Flux Schnell), VAEs, samplers and resolution settings ⦁ LoRA: loading, weights tuning and training basics ⦁ Video generation: AnimateDiff, SVD or comparable pipelines ⦁ Building API end points around ComfyUI for programmatic generation ⦁ Auto scaling GPU workers (RunPod serverless or similar) and cost efficient scaling to zero ⦁ GPU environments: RunPod, Vast.ai, local CUDA or cloud GPU instances It would be better if you also know Automatic1111, Kohya for LoRA training, upscaling pipelines, FastAPI, queue based generation systems, or node based creative tools like Weavy (Figma Weave). How we'll work: This is a $100 fixed price project with a clear deliverable list above. You will get access to the GPU environment and a Slack channel for quick communication. If the result is good, there is follow up work: LoRA training on our brand assets, deeper integration of the API end points into our product, and ongoing generation pipelines. How to apply: Skip the generic pitch. Start your proposal with the word "Canvas" so I know you actually read this, then share 2 or 3 sample outputs (images or videos) generated with your own ComfyUI workflows, and tell me which GPU environment you would recommend for us and why. Real outputs beat certificates.