17:11
Photoshop
Ten Photoshop workflows that used to take retouchers hours or years of skill to master — object removal, canvas expansion, shadow/lighting fixes, reflection cleanup, background removal, compositing accessories, distraction removal, backdrop cleanup, deblurring, and upscaling — can now be done in minutes with AI tools inside and outside Photoshop, but the presenter argues core Photoshop/design fundamentals remain necessary to catch and fix AI's frequent, subtle mistakes.
AI has made many previously time/skill-intensive Photoshop tasks fast (minutes instead of hours), especially for non-creative cleanup work.
Generative Fill removal/replacement (e.g., removing a person, glasses) often subtly alters untouched details like eyes, lips, or skin texture, requiring a negative-mask trick to localize the AI edit to only the needed area.
Generative Expand handles simple, blurred backgrounds well but produces garbled, low-quality results ('AI mumble rap') on complex backgrounds, and Photoshop doesn't let users apply better models like Nano Banana to it.
Workaround for bad expansion: add a black background layer, export, and run the expansion through an external tool (Magnific with Nano Banana 2/Pro, or Adobe Firefly) which supports higher resolution (4K).
AI can remove shadows from a face or an entire scene via a text prompt in Generative Fill, and can even change ambient lighting (e.g., 'change the lighting to overcast and soft') while mostly preserving the subject.
AI still struggles with stylized creative lighting (e.g., a three-point flash setup with rim light), altering the eyes, teeth/braces, and skin texture — this is framed as the boundary where AI needs improvement.
Camera Raw Filter's Remove > Reflections feature automatically cleans reflections off glass surfaces across an image, with adjustable strength and a 'Best' quality setting, but it does not yet remove glare from glasses (which still needs Generative Fill).
Remove Background now handles complex object selections well, but understanding blend modes (duplicating the layer, setting Multiply) is still needed to fix edge artifacts like hair picking up a same-color background cast.
Generative Fill's reference-image option lets users composite a specific real object (jewelry, sunglasses, a watch) onto a subject by uploading a photo of it and prompting 'add that [object]', replacing what used to require years of dodge/burn/warp compositing skill.
AI compositing from a reference image is not always accurate — a watch generated onto a wrist appeared worn backwards with garbled text, prompting the presenter to say he'd never buy a product based on an AI-generated product image.
The Remove tool's 'Find Distractions' feature automatically detects and removes categories like wires/cables, people (including the photographer's own shadow), dirt, poles, dust, text, and general visual clutter.
The Retouch4Me 'Clean Backdrop' plugin (used by the presenter for 5 years) automates backdrop wrinkle/imperfection cleanup that used to be manual retouching work.
Generative Fill can extend a backdrop (e.g., 'extend the backdrop all the way to the ground') in place of older manual stretch/clone techniques from an earlier tutorial video.
AI-based prompted fill ('fix and sharpen the blurry nose and the lips') can reconstruct out-of-focus image areas far better than the older Unsharp Mask sharpening filter ever could.
Magnific's AI image upscaler (Precision model) can dramatically increase resolution and generate convincing fine detail/texture (e.g., wood grain) not present in the original low-resolution photo.
The presenter's closing claim: people with strong Photoshop/design/photography fundamentals stay ahead of people who 'just know AI,' because AI-only users lack 'taste' and don't know what to do when AI output goes wrong.
Generative Fill (object removal/replacement) — A Photoshop AI feature that fills a selected area based on a text prompt, used to remove or replace objects like a person in the background, glasses, or braces. Apply: Make a marquee selection around the object, click Generate a Fill, type a prompt like 'remove the man in the background,' pick a model, and click Generate.
Negative mask touch-up trick — A masking technique for constraining an AI-generated fill to only the pixels that actually need to change, since Generative Fill can subtly alter untouched details like eyes or skin texture. Apply: Drag the AI edit's mask to the trash, Alt/Option-click the mask button to create a negative (black) mask, then paint with a soft white brush only over the specific areas that need the edit.
Generative Expand — A Photoshop AI feature that extends the canvas and fills the new area with AI-generated content matching the scene. Apply: Press C for the Crop tool, set an aspect ratio and drag the canvas larger, then click Generative Expand, optionally typing a prompt and choosing a model such as Firefly Fill and Expand, and click Generate.
Black-background workaround for external expand/upscale tools — A workaround for expanding complex backgrounds that Generative Expand renders as 'garbage,' since better models like Nano Banana aren't available inside Photoshop's Generative Expand and don't read transparent pixels. Apply: Add a black solid-color adjustment layer beneath the layer to fill the transparent expanded area, export the file, then upload it to Magnific or Adobe Firefly and prompt something like 'remove the black bars on both sides' at the desired aspect ratio and resolution.
AI shadow/lighting removal via Generative Fill — Using a text prompt through Generative Fill (or an external tool with Nano Banana) to remove shadows from a face or subject, or to change the lighting mood of an entire scene. Apply: Select the image or subject area, click Generate a Fill, and type a prompt such as 'remove all shadow from the face' or 'remove all shadows and change the lighting to overcast and soft,' choosing a model like Firefly Image 5.
Remove Reflections (Camera Raw Filter) — A built-in Camera Raw 'Remove' feature for cleaning up reflections in glass surfaces across an image, with an adjustable strength and quality setting. Apply: Duplicate the background layer, convert for Smart Filters, open Camera Raw Filter, go to the Remove section, scroll to Reflections, click Apply, and adjust the amount slider and quality (set to Best).
Remove Background + Multiply blend-mode edge fix — An AI subject-selection tool paired with a manual blend-mode trick to fix hair edges that pick up color cast from a new same-colored background. Apply: Click Remove Background, duplicate the cut-out layer, set the duplicate's blend mode to Multiply to darken the edge halo, turn the original layer back on, and erase around the edges on its mask with a black brush to blend the two.
Generative Fill with reference image — A Generative Fill mode that accepts a reference image (e.g., jewelry or a watch) so the AI composites that specific object onto the subject instead of inventing one from text alone. Apply: Select the target area, open Generative Fill, choose a model like Nano Banana 2, click the reference-image option, upload the reference photo, and type a prompt like 'add that jewelry.'
Remove tool — Find Distractions — A Photoshop AI feature that automatically detects and lets you remove categories of unwanted elements: wires/cables, people, dirt, poles, dust, text, and general urban clutter. Apply: Create a new layer, select the Remove tool, choose 'Find Distractions' and a category (e.g., wires and cables, or people), ensure 'Sample All Layers' is checked, and click Remove.
Retouch4Me 'Clean Backdrop' plugin — A third-party AI Photoshop plugin, part of a Retouch4Me suite the presenter has used for 5 years, that automatically smooths and cleans backdrop wrinkles and imperfections. Apply: Duplicate the layer (Ctrl/Cmd+J), go to Filter > Retouch4Me > Clean Backdrop, choose a setting like 'bigger dirt,' make the mask, and hit Apply to get the fix isolated on a new layer.
Generative Fill for backdrop extension — Using Generative Fill to extend a photo's backdrop outward or downward instead of manually stretching or cloning it. Apply: Select the empty area to fill, type a prompt like 'extend the backdrop all the way to the ground,' generate, then select the resulting mask and fade the edit in with a black brush for a smoother blend.
Prompted AI deblurring (replacing Unsharp Mask) — A Generative Fill prompt-based method for reconstructing blurry, out-of-focus image areas, contrasted with the older Unsharp Mask sharpening filter that could only fake detail by increasing radius. Apply: Instead of Filter > Sharpen > Unsharp Mask, select the blurry area and use Generative Fill with a prompt like 'fix and sharpen the blurry nose and the lips.'
Magnific AI Image Upscaler (Precision model) — An external paid AI upscaling tool used to dramatically increase image resolution and generate realistic fine texture detail (e.g., wood grain) beyond what the original low-resolution photo contained. Apply: Open the Magnific image upscaler, upload the photo, choose the Precision model, set the scale factor (e.g., 4x), disable extra sharpness/grain if undesired, and click Upscale.
The recurring 'negative mask' trick (delete the AI edit's mask, Alt/Option-click to create an inverted black mask, then paint white only where change is wanted) is presented as the actual surviving skill — not avoiding AI, but constraining it precisely, which itself requires traditional Photoshop masking literacy.
Photoshop appears to deliberately restrict which AI models are available to specific tools (e.g., Nano Banana isn't offered for Generative Expand), which is why the presenter routes through external platforms like Magnific — a product/licensing constraint as much as a technical one.
AI output is shown to be inconsistent even across repeated attempts with the same prompt on the same image (e.g., regenerating glasses removal three times gave a shrunk face, then a good result, then a shifted mouth), implying iterative regeneration plus manual masking is treated as a normal step of the workflow rather than a fallback.
The video implies a moderation-bypass dynamic: directly asking the AI to generate a 'no shadow' version of someone's face 'would probably ban me,' but reaching the same visual result by framing the prompt as a removal/fill task inside Generative Fill succeeds.
The presenter draws an explicit line between 'non-creative' processes (distraction, reflection, and shadow removal) where he treats AI as functionally sufficient, and 'creative' processes (stylized lighting setups) where AI still visibly degrades identity and texture — this line, not a blanket AI-good/AI-bad stance, structures the whole video.
«That's not art. That's AI mumble rap. This is crazy bad right there.»
— 00:10
«But it's the difference between spending 4 hours and 4 minutes.»
— 00:29
«Remember, black hides.»
— 01:42
«Keep in mind, AI is not perfect. It took me three tries for it to happen.»
— 11:45
«And I would personally never buy a product based on an AI image of the product.»
— 12:22
«You're always leaps and bounds ahead of people who just know AI if you have the core fundamental Photoshop skills or core fundamental design or photography skills.»
— 16:25
«Someone who doesn't understand it may create something with AI but they won't have taste and when AI goes wrong they would not know what to do.»
— 16:33
Reception
Audience reception is solidly positive, with strong appreciation for the balanced, educational approach to AI; however, there's meaningful concern about prioritizing efficiency over craft and persistent anxiety about job displacement.
A demonstration-heavy tutorial that backs its 'AI made this obsolete' framing with paired before/after examples across ten workflows, while its own on-screen failures (garbled expanded backgrounds, a backwards AI-generated watch) double as evidence for its closing argument that AI output still needs a skilled hand to catch and correct it.
17:11