After publishing my RunPod review and pricing breakdown, the question I get most is simple: how do you actually go from a blank RunPod account to a finished product photo?
This is the step-by-step walkthrough I wish someone had handed me the first time I tried it, written specifically for ecommerce sellers who want usable product imagery, not AI art for its own sake.
You do not need to be a developer to follow this. You do need about an hour for your first session and a willingness to click around a slightly intimidating node-based interface. After that first hour, generating a new batch of product images takes ten to fifteen minutes.
Step 1: Create Your RunPod Account and Add Credit
Sign up and add a small amount of credit, $10 to $20 is more than enough to experiment for your first several sessions. RunPod bills by the second, so you are not committing to a subscription, just funding a prepaid balance you can top up whenever you need more.
Step 2: Deploy the ComfyUI Template
From your dashboard, choose Deploy and search the template library for a ComfyUI template. RunPod maintains a ComfyUI with Flux template that installs ComfyUI, the Flux.1 image model, ComfyUI Manager, and a set of default workflows automatically, so you are not manually installing dependencies from scratch. Select an RTX A5000 or RTX 4090 GPU for this first run, both run well under a dollar an hour and handle product image generation comfortably.
Click deploy, wait a couple of minutes for the pod to spin up, then open the ComfyUI web interface directly from your RunPod dashboard. You will land on a node-based canvas that looks intimidating at first but is really just a flowchart: each box is a step, and lines connect the output of one step to the input of the next. RunPod’s own ComfyUI documentation walks through the deployment screen in more detail if you want a second reference alongside this guide.
If the default template does not include the specific nodes referenced later in this walkthrough, ComfyUI Manager, which comes preinstalled on RunPod’s template, lets you search for and install missing custom nodes directly from inside the interface rather than editing any configuration files by hand.
Run Your Own AI Image Models for Pennies an Hour
Deploy a prebuilt ComfyUI template and start generating product images with GPUs starting around $0.27/hr.
Step 3: Understand the Four-Stage Product Photo Workflow
Most ecommerce product photography workflows in ComfyUI follow the same basic four-stage chain, regardless of which specific tool combination you use. Understanding these four stages makes the whole process click, even if the exact node names vary between templates.
Background removal. You start with a clean photo of your product, ideally shot on a plain background. A background removal model, commonly BiRefNet in current ComfyUI workflows, isolates the product from its original background and produces a transparent cutout.
Scene generation. Next, an image generation model, either Stable Diffusion or Flux depending on your template, generates a new background scene based on a text prompt describing the setting you want, a kitchen counter, an outdoor patio, a minimalist studio backdrop, whatever fits your brand.
Relighting and compositing. The product cutout gets placed into the new scene and relit to match the scene’s lighting direction and color temperature. A relighting model called IC-Light handles this step in most current workflows, and it is the difference between a product that looks pasted on and one that looks like it was actually photographed in that setting. Community-published workflow breakdowns on Civitai are a useful reference if you want to see how other users have chained these nodes together for different product categories.
Batch export. Once you have a result you like, ComfyUI can batch-generate variations automatically, cycling through different backgrounds, angles, or lighting setups so you end up with several usable options from a single session instead of one image at a time.
Step 4: Upload Your First Product Photo
Take a clean photo of your product on a plain white or neutral background using nothing fancier than a smartphone camera and decent lighting. Upload it into the workflow’s load image node. The cleaner and better-lit this source photo is, the better every downstream step performs, so do not skip putting real effort into this first shot even though the rest of the process is automated.
Run the background removal step first and confirm the cutout looks clean before moving forward. If the edges look rough or part of the product got cut off, that is worth fixing at this stage rather than compounding the problem through the rest of the pipeline.
Step 5: Write Your Scene Prompt
This is the step that determines most of the final image quality. Describe the setting specifically: lighting conditions, surface material, time of day, and mood. “Product on a marble countertop near a window, soft morning light, minimalist” produces a dramatically different (and usually better) result than just “product on a table.”
Generate a handful of variations at low resolution first to find a prompt and seed you like, then commit to a full-resolution render only once you have something worth spending the extra GPU time on. This two-pass approach is the single biggest time and money saver once you get comfortable with the workflow.
Step 6: Fine-Tune Lighting and Composition
Once the base composite looks right, the relighting stage lets you adjust the direction and intensity of light hitting your product to better match the scene. This is where a lot of the “does this actually look real” quality comes from, and it is worth spending a few extra minutes here rather than accepting the first pass.
If your product has reflective surfaces, glass, metal, or glossy plastic, pay extra attention to this step, since mismatched reflections are the fastest way for an AI-composited image to look obviously fake to a shopper.
Step 7: Batch Generate Variations
Once you have a workflow producing results you are happy with, use ComfyUI’s batch generation to run the same pipeline across multiple background prompts or product photos in one queued run. This is where the real time savings show up: instead of babysitting each image individually, you queue ten or twenty variations and come back once they have finished rendering.
For a typical high-ticket product catalog, I recommend generating three to five background variations per product: one clean studio shot, one or two lifestyle scenes, and one detail or close-up angle. That gives you enough variety for a product page gallery without spending an entire afternoon on a single SKU.
Step 8: Download and Organize Your Results
Download every image you plan to use before shutting down your pod. RunPod’s persistent storage has not always reliably survived pod restarts according to user reviews, so treat “download now” as a hard rule rather than an optional step, regardless of which storage tier you are on.
Organize downloads by product SKU as you go rather than after the fact. A folder structure that mirrors your store’s product catalog saves real time when you are uploading images back into Shopify or whatever platform you sell on.
Step 9: Save Your Workflow for Next Time
Once you have a working pipeline, export and save the ComfyUI workflow file locally. The next time you rent a pod, you load this saved workflow instead of rebuilding it from the default template, which turns your first-session hour into a ten or fifteen minute session going forward.
Keep a couple of workflow variants saved, one for clean studio-style product shots and one for lifestyle scene compositing, since these serve different purposes across a product listing and swapping between saved workflows is faster than adjusting one workflow’s settings back and forth.
Step 10: Shut Down the Pod
Once you have downloaded everything, stop or terminate the pod. RunPod bills for GPU time until you do this manually, and this is the single most common way new users end up with a bill larger than expected. I cover the exact cost breakdown, including this specific gotcha, in my RunPod pricing guide.
Choosing Between Stable Diffusion and Flux
Most current ComfyUI templates default to either Stable Diffusion or Flux for the scene generation step, and the choice matters more than it might seem. Flux, documented on Hugging Face’s model page, generally produces more photorealistic results out of the box with less prompt tweaking, which makes it a better starting point if you are new to this and do not want to spend hours refining prompts to get a usable image.
Stable Diffusion has a larger ecosystem of community fine-tunes and add-ons built up over a longer period, which can be an advantage if you find a specialized model trained specifically for product photography or a particular aesthetic your brand wants. For a first attempt, though, stick with whatever model your chosen template defaults to rather than switching mid-workflow, since troubleshooting two variables at once, a new model and a new workflow, makes debugging unnecessarily hard.
Building a Repeatable Weekly Routine
Once you have a saved workflow, the most efficient pattern for an ongoing store is a weekly or biweekly batch session rather than generating images one product at a time as they get added. Queue up source photos for every new or updated product since your last session, run them through your saved pipeline in one sitting, and download everything at once before shutting the pod down.
This batching approach also makes the GPU cost trivial relative to the value delivered, since you are spreading one pod session’s cost across potentially a dozen or more product images rather than spinning up a fresh pod for every single item. I break down the real dollar cost of sessions like this in my RunPod pricing guide.
Common Mistakes to Avoid
Skipping the low-resolution test pass is the most expensive mistake, since it means paying full GPU rates to discover a prompt does not work the way you expected. Test cheap, commit to full resolution only once you like the result.
Using a poorly lit or blurry source photo is the second most common issue. No amount of downstream AI processing fixes a bad starting image, so put genuine effort into your original product photo even though everything after that is automated.
Forgetting to save your workflow file is the third, and it is the one that costs you the most time rather than money, since you end up rebuilding a working pipeline from scratch on your next session instead of loading a saved one.
A fourth mistake worth naming: trying to perfect every single image before moving to the next product. Generate your batch, pick the best two or three results per product, and move on. Chasing a flawless render on every single SKU burns far more GPU time and personal patience than the marginal quality improvement is usually worth for a working product listing.
When This Workflow Is Not the Right Fit
If you need a single polished image today and have zero appetite for a technical learning curve, a subscription AI photography tool will get you there faster on your first attempt, even though it costs more over time. This ComfyUI and RunPod approach pays off once you are generating images regularly enough that the time investment in learning the workflow gets amortized across dozens of future sessions.
It is also worth prioritizing correctly if your store is still early. Get your high-ticket dropshipping fundamentals and supplier relationships locked down before investing hours into a custom AI imaging pipeline, since polished product photos do not matter much if you do not have reliable products to photograph yet, and no amount of great imagery rescues a store with a shaky supplier foundation underneath it.
Frequently Asked Questions
Do I need design or Photoshop experience to follow this?
No. The workflow handles background removal, scene generation, and relighting automatically once configured. Your job is mostly writing good prompts and picking the results you like.
How long does the whole process take once I am set up?
After your first session, generating a new batch typically takes ten to fifteen minutes of active work plus rendering time, most of which runs unattended once you queue a batch.
What if my product has a complicated shape or reflective surface?
Give extra attention to the relighting step and expect to run a few more test passes. Glass, metal, and glossy plastics are the hardest surfaces for any AI compositing workflow to get convincingly right.
Can I use this for lifestyle shots, not just clean product photos?
Yes, that is actually where this workflow shines most, since generating varied lifestyle scenes around a single product photo is far cheaper than booking multiple physical photo shoots in different settings.
Should I use Community Cloud or Secure Cloud for this?
Community Cloud is fine for most sellers generating images from their own already-public product photos. I break down when Secure Cloud is worth the extra cost in my RunPod review.
What resolution should my final product images be?
Match whatever your ecommerce platform recommends for product listings, typically at least 2000 pixels on the longest side, and generate at that resolution only for your final chosen variations to save GPU time on test passes.
Can I automate this so new products get images generated automatically?
Yes, RunPod’s Serverless option can trigger this workflow automatically when a new product gets added to your catalog, though that setup is more advanced than the manual Pod-based workflow covered here and worth attempting only after you are comfortable running the manual version reliably.
Our Services
If you want direct help building or scaling a store, I offer 1-on-1 coaching, a done-for-you store build service, and a full turnkey store package for people who want to skip the setup phase entirely. I also run supplier recruiting, Google Shopping ads management, and SEO services for stores that are ready to scale traffic.
Free Resources
If you are just getting started, grab my beginner’s guide, browse the free resource library, check out the blog for more breakdowns like this one, or join my Patreon community for ongoing support.
Before you invest an afternoon into this workflow, make sure your product sourcing is solid. My complete guide to finding suppliers is the place to start if you have not locked in your catalog yet.
And once your store is generating real revenue, my walkthrough on business formation for ecommerce founders covers the legal and financial foundation worth having in place as you scale up content production.
Not Sure Which Niche to Build Your Store Around?
Get my full list of proven high-ticket niches so you know exactly where to focus before you pick a single product.
Related Articles
AI Product Photography for Ecommerce
Best AI Product Photography Tools for Ecommerce in 2026
What Is High-Ticket Dropshipping
Trevor Fenner is an ecommerce entrepreneur and the founder of Ecommerce Paradise, a platform focused on helping entrepreneurs build and scale profitable high-ticket ecommerce and dropshipping businesses. With over a decade of hands-on experience, Trevor specializes in high-ticket dropshipping strategy, niche and product selection, supplier recruiting and onboarding, Google & Bing Shopping ads, ecommerce SEO, and systems-driven automation and scaling. Through Ecommerce Paradise, he provides free education via in-depth guides like How to Start High-Ticket Dropshipping, advanced training through the High-Ticket Dropshipping Masterclass, and fully done-for-you turnkey ecommerce services for entrepreneurs who want a faster, more hands-off path to growth. Trevor is known for emphasizing sustainable, real-world ecommerce models over hype-driven tactics, helping store owners build scalable, sellable, and location-independent brands.
