21:06
Adobe Photoshop
Adobe's generative AI credit system in Photoshop and Lightroom follows one general rule — creating new image pixels costs credits, while selections and detection-based tools do not — but the video argues the system is still confusing and, in the presenter's hands-on test, most credit-costing generative-addition features weren't worth their cost.
General rule: generating new image pixels costs credits; selections, detection, and denoising do not.
Credit allowance depends on the user's Creative Cloud plan; credits can be purchased separately or gained via plan upgrade.
Credits reset monthly and do not roll over; once spent, they cannot be recovered even if the result is rejected.
Free AI tools: select subject, remove background, object selection tool (people/car/hat/sky), select sky, AI masks in Camera Raw/Lightroom (subject, sky, background, landscape sub-masks), lens blur.
Generative expand extends canvas borders via Firefly model 3 at 1 credit per variation.
Generative fill removes or adds objects by generating pixels; an empty prompt auto-detects removal intent; Firefly costs 1 credit per generation.
The remove tool's cost depends on its AI toggle: 'auto' may invoke AI when needed, 'generative AI on' charges credits, 'generative AI off' is free and falls back to traditional content-aware fill.
Recommended workflow: try the remove tool with no AI first, and only escalate to AI (1 credit) if the non-AI result isn't good enough, versus spending more otherwise.
Partner models (non-Adobe) can substitute for Firefly inside generative fill, each with its own credit cost that is hidden until the user hovers over the properties icon on the generative layer.
Partner model pricing: Firefly models 1 and 3 = 1 credit/variation; Flux Context Pro = 10 credits/generation; Flux 2 Pro = 20 credits/generation (single variation only, unlike Firefly's 3); Gemini 2.5 with Nano Banana = 10 credits/generation; Gemini 3 with Nano Banana Pro = 40 credits/generation.
Partner model pricing is stated to be subject to future changes.
A test of adding a bunny to a photo across three different models consumed 73 credits total.
Credit balance can be checked via Adobe Creative Cloud > 'your plan' > 'view your generative AI balance', or directly via the Adobe account.
The presenter started the tutorial at 4,000 credits and ended at 3,810, spending 190 credits total to produce the video.
The entire first half of the video (non-generative tools) used zero credits; the entire second half (generative tools) used all 190 credits spent.
The presenter's honest assessment: only generative expand (crop) and the remove tool were genuinely valuable; every other generative-addition attempt didn't produce usable results.
Select Subject — A cloud-based AI selection tool that automatically selects the main subject in an image, using AI but not generating new pixels. Apply: Click 'select subject' in the contextual taskbar on a layer to get an automatic subject selection at zero credit cost.
Remove Background — An AI tool that isolates and removes the background around a subject without generating new pixels. Apply: Click remove background to instantly cut out the background at no credit cost.
Object Selection Tool — An AI-powered tool that lets users select specific object categories (people, cars, hats, sky, etc.) within a photo without using credits. Apply: Choose the object selection tool, then pick a category like 'select people' to select an entire person or specific parts of them.
Select Sky — An AI tool dedicated to isolating the sky region of an image, at no credit cost. Apply: Use select sky to isolate the sky for targeted edits like color or exposure adjustments, independent of the rest of the image.
AI Masks (Camera Raw / Lightroom) — A native masking system using AI to generate subject, sky, background, and landscape sub-masks (sky, vegetation, natural ground) without using credits. Apply: Open the masking panel in Camera Raw or Lightroom and select a mask type (e.g. landscape > sky) to build free, non-destructive local adjustments.
Lens Blur — An AI-driven effect that applies background blur to simulate shallow depth of field, without generating new pixels and without using credits. Apply: Apply the lens blur filter to a photo to add background blur for free.
Generative Expand — A canvas-extension feature powered by Firefly model 3 that generates new pixels beyond the original image borders, costing 1 credit per variation. Apply: Use the crop tool's fill option or generative expand to extend a canvas's edges, expecting to spend 1 credit for each of up to 3 variations generated.
Generative Fill — A tool that removes or adds objects to an image by generating new pixels via a selectable AI model, at a cost that depends on the chosen model. Apply: Make a selection with the selection brush, leave the prompt empty to auto-detect object removal or type a prompt to add content, then pick a model before generating.
Remove Tool — A brush-based object-removal tool whose credit cost depends on an AI toggle set to auto, on, or off. Apply: Paint over an unwanted object with 'generative AI off' first (free, uses content-aware fill); if unsatisfactory, switch to AI mode to spend 1 credit for a single-variation result on a new layer.
Content-Aware Fill — The traditional, non-AI pixel-fill algorithm the remove tool falls back to when its AI toggle is off, costing no credits. Apply: Rely on it as the free first-resort for object removal before escalating to the AI-powered remove tool.
Firefly Image Model 1 / Model 3 — Adobe's native generative models used for generative fill and expand, priced at 1 credit per variation. Apply: Select Firefly model 1 or 3 from the model dropdown in a generative layer for the cheapest per-variation generations.
Flux 2 Pro — A partner generative model available inside generative fill, priced at 20 credits per generation and limited to a single variation. Apply: Select Flux 2 Pro from the model dropdown for a single-shot generation, understanding it costs far more per attempt than Firefly.
Flux Context Pro — An older partner Flux model available inside generative fill, priced at 10 credits per generation. Apply: Choose Flux Context Pro as a mid-cost partner-model alternative to Flux 2 Pro when generating fills.
Gemini 3 with Nano Banana Pro — Google's latest partner model integrated into Photoshop's generative fill, priced at 40 credits per generation and noted as best for full-image regeneration. Apply: Select this model when attempting a full-image regeneration workflow, accepting the highest per-generation credit cost among the tested models.
Gemini 2.5 with Nano Banana — An earlier Google partner model available in generative fill, priced at 10 credits per generation. Apply: Choose it as a cheaper Google-model alternative to Gemini 3 Nano Banana Pro.
Checking Generative AI Balance — A method for viewing remaining AI credits and premium-feature balance through Adobe Creative Cloud or an Adobe account. Apply: Go to Creative Cloud > 'your plan' > 'view your generative AI balance' (or the Adobe account directly) to track credits spent and remaining.
Partner-model credit cost is not disclosed upfront in the interface — it only surfaces if the user manually hovers over the properties icon on the generative layer, meaning users can commit to a costly model without seeing the price first.
Credit cost and output quality are decoupled in the presenter's test: paying more (e.g. 40 credits for Gemini 3 Nano Banana Pro) did not reliably produce a usable result when adding new content to an existing photo, and none of the tested addition attempts across models were judged worth using.
The single costliest use case in the video (adding an object to an existing photo) was also rated the least valuable, while the cheapest AI-assisted use cases (canvas expansion, object removal) were rated the most valuable — inverting a naive 'you get what you pay for' assumption.
The remove tool's non-AI path (content-aware fill) is explicitly framed as a free first-resort, with AI escalation reserved for failures, turning credit management into a deliberate two-step workflow rather than a default AI-first habit.
The gap between the cheapest and priciest models for functionally similar tasks is up to 40x (1 credit Firefly vs. 40 credits Gemini 3 Nano Banana Pro), a spread the video treats as inherently confusing to end users.
«AI is not free. Every time you create a generation, it's going to be using some credits»
— 00:03
«Basically, a general guideline is that if you're creating new image pixels, you're going to be using credits. If you're just doing things like creating selections or denoising, those things won't use any credits at all.»
— 00:26
«These do not roll over, they reset every single month, and credits, once they're used, they cannot be undone, even if you don't like the results.»
— 00:55
«if you want to do a removal, start with the remove tool with no AI. If it didn't do a good enough job, then use the remove tool with AI and you're going to see it's going to use one credit instead of three.»
— 10:55
«Adobe introduced these partner models because AI models are incredibly complex and expensive to create.»
— 12:51
«what did I use? One, two, three credits plus 10, so 13 plus 20, 33, 73 credits I just used to create those bunnies.»
— 18:19
«it's just kind of a crap shoot. You're going to spend a ton of credits and maybe not get results that are even usable whatsoever.»
— 18:52
«I used 190 credits to create today's video, which I'm okay with.»
— 19:52
«the only things that were really actually that value was expanding my crop and removing my subject. Everything else that I used wasn't even like it didn't even get me anywhere.»
— 20:03
Reception
Audiences appreciate the tutorial's clarity and honesty but are deeply frustrated with Adobe's credit-based pricing model, with many questioning the company's direction and considering alternatives.
PHLEARN's video maps Adobe's credit system claim-by-claim with specific timestamps, tool names, and costs, and is transparent about its own methodology (checking balance before/after); its central limitation as a source is that it documents pricing the video itself flags as subject to change, and its 'value' judgments (expand/remove good, additions bad) are a single presenter's subjective read on one test session.
21:06