06:47
AI product photography
Thomas Lundström presents his personal, tool-chained workflow for AI-assisted product photography: use ChatGPT to brainstorm and generate Midjourney-formatted prompts, generate the background image in Midjourney first, then photograph the real product to match that background, and finally blend the two in Photoshop (Neural Filters) and Lightroom — letting a small creator produce polished product shots without a location shoot or a big studio budget.
Start by using ChatGPT's 'Mid Journey MJ Prompt Generator V6' GPT (found under Explore GPTs) to turn a plain description of the desired shot into five Midjourney-ready prompts.
Paste the ChatGPT-generated prompts into Midjourney's Discord to generate candidate background images.
Go back and forth between ChatGPT and Midjourney, refining the prompt and regenerating, to fine-tune the background's look — described as the most critical step in the workflow.
Generate the AI background before shooting the real product, not the reverse, because matching a background to an already-shot product is much harder than matching a shot to an existing background.
Because the background is generated by AI, the physical studio setup can stay simple: one light and a neutral wall.
Try to find a physical surface similar to the one visible in the AI-generated image, to make the later Photoshop blend easier.
Use Sony's Imaging Edge Desktop app to overlay the AI-generated background image on the camera's live viewfinder, aligning perspective and zoom in real time while shooting.
Take an additional product photo lit with an orange light on the back of the bottle to replicate a glow effect present in the AI background.
Editing order: base grade in Lightroom, then blend in Photoshop, then final adjustments back in Lightroom.
In Photoshop, use Neural Filters' Super Zoom to add detail and upscale the product photo to match the AI background's resolution.
Use Select Subject to isolate the product and remove its background, applying the same selection to both the normal and glow-lit layers.
Use a soft (hardness 0) Eraser brush to reveal the glow layer underneath before placing the product into the AI background.
Use Neural Filters' Harmonization, referencing the AI background layer, to match the product's colors to the background; lower the default 75 strength (around 50) and adjust brightness for a more natural result.
Remove the original product from the AI-generated background and use Generative Fill to help blend the real product's shadow into the scene.
Finish in Lightroom with the Auto adjustment, slider tweaks, and grain applied across the whole composited image to help the two sources blend together.
The claimed payoff: a small creator can produce compelling product photography without traveling to a location or funding a big set — just a camera, some lights, and imagination.
MJ Prompt Generator V6 (ChatGPT GPT) — A custom GPT in ChatGPT's Explore GPTs tab that generates five Midjourney-formatted prompts from a plain-language product photo description. Apply: Open the 'Mid Journey MJ Prompt Generator V6' GPT, describe the desired product photo (subject, framing, setting, focus), and use the five returned prompts directly in Midjourney's Discord.
Midjourney background generation — An AI image generator used via Discord to turn ChatGPT-produced prompts into candidate background images. Apply: Paste the ChatGPT-generated prompts into Midjourney's Discord and iterate with ChatGPT on the wording until a background with the desired composition (framing, blur, glow elements) is produced.
ChatGPT–Midjourney iteration loop — A back-and-forth refinement process where results from Midjourney inform new prompt requests to ChatGPT. Apply: After reviewing Midjourney outputs, return to ChatGPT to adjust and regenerate prompts until the background's look matches what's wanted; described as the most critical step for getting the right look.
Background-first shooting order — A rule that the AI background must be generated before the physical product is photographed, rather than the reverse. Apply: Generate the AI background image first, then set up the physical shoot to match its perspective, zoom, and lighting, rather than creating a background to match an already-taken product photo.
Imaging Edge Desktop viewfinder overlay — A Sony camera companion app used here to overlay a reference image (the AI-generated background) onto the live camera viewfinder. Apply: Load the AI background image into Imaging Edge Desktop and use it as an overlay on the live viewfinder to match the camera's perspective and zoom while shooting the product.
Secondary glow-light photo — A supplementary product photo shot with an orange light aimed at the back of the bottle to mimic the ambient/backlit glow present in the AI background. Apply: After the main product shot, take an additional photo with a colored (orange) light hitting the back of the product, then later blend this glow layer into the main image with a soft eraser.
Lightroom–Photoshop–Lightroom editing order — A three-stage editing pipeline order used to combine the AI background and the real product photo. Apply: Apply an initial base color grade to the product photo in Lightroom, move to Photoshop to blend it into the AI background, then return to Lightroom for final adjustments.
Neural Filters — Super Zoom — A Photoshop Neural Filters feature that adds detail and upscales an image's resolution. Apply: Apply Super Zoom to the product photo to enlarge it and add detail so its resolution matches the AI-generated background image.
Select Subject — A Photoshop tool that automatically selects the main subject in a photo to isolate it from its background. Apply: Use Select Subject on the product photo to select the bottle, refine as needed, and apply the same selection to remove the background from both the normal and glow-lit layers.
Soft Eraser (hardness 0) glow blending — A manual masking technique using a zero-hardness eraser brush to reveal parts of a layer underneath. Apply: With the glow-lit layer beneath the main product layer, use a soft (hardness 0) Eraser brush on the top layer to gradually reveal the glow from behind before placing the product into the AI background.
Neural Filters — Harmonization — A Photoshop Neural Filters feature that color-matches a foreground layer to a chosen reference layer. Apply: Select the AI background layer as the harmonization reference for the product layer, then lower the default strength (75) to around 50 and adjust brightness for a more natural color match.
Generative Fill — A generative AI fill tool used to remove and regenerate parts of an image. Apply: Use Generative Fill to remove the original product from the AI-generated background and to help blend the real product's shadow into the background more naturally.
Lightroom Auto + grain finishing — Lightroom's automatic adjustment feature combined with an added film-grain effect applied across the whole composited image. Apply: In the final Lightroom pass, apply the Auto adjustment, tweak sliders as needed, and add grain over the entire image so the AI background and real photo elements visually blend together.
The video frames shooting order as the linchpin of the whole method: generate the background first, then shoot to match it in-camera, because reverse-engineering a background to fit an existing photo is claimed to be far harder.
Repurposing Sony's Imaging Edge Desktop (a tethering/remote-shooting app) as a live viewfinder overlay against a static AI image is an improvised alignment technique not designed for this use.
Deliberately shooting a second 'glow' photo under colored (orange) light, then hand-blending it in with a soft eraser, is used to physically fake ambient/backlight that doesn't exist in the real studio scene.
Matching the physical prop surface to the texture seen in the AI-generated image is called out explicitly as easing the Photoshop blend, not just as a styling choice.
The default strength (75) of Photoshop's Harmonization Neural Filter is explicitly flagged as too strong for natural results; the creator recommends dropping it to around 50 with brightness tweaks.
Adding grain across the entire final composite is used specifically as a blending trick to unify the AI-generated and camera-captured elements, not purely for stylistic grain.
«in this video I'm going to show you my workflow of creating product photos like these using AI»
— 00:00
«this way around because if you take photos of your product first and then try to create backgrounds that match your photos it's going to be very hard to get that to look similar but if you generate the backgrounds first first you can in camera try to really match up the shot so it fits into the background»
— 02:18
«of course and I tried to find a similar surface to the one that was in the AI generated image because this was will help when we try to blend the two images together in Photoshop»
— 02:53
«this over here is kind of the superpower of this workflow so I use a Sony camera and Sony has an app called Imaging Edge desktop»
— 02:59
«some tweaks we can get it to match even Closer by default it's at 75 and I think that's usually a bit too strong so I like to play around with the brightness and maybe drop the strength down to somewhere around 50 in order to get a look that is more natural to the photo and the environment»
— 05:20
«I think this is a cool way just for a small Creator to create something that doesn't require you to go to a location or a big budget for a big set you just need your camera some lights and your imagination»
— 06:19
Reception
Audience found the tutorial clear and genuinely helpful, with strong gratitude, though some note practical limitations and newer alternatives have since emerged.
The video documents a genuinely manual, multi-tool pipeline (ChatGPT, Midjourney, a Sony DSLR, Photoshop Neural Filters, Lightroom) rather than a one-click solution, laying out each step and specific setting (e.g., Harmonization strength) in enough detail to be reproduced.
06:47