Correction, 2026-10-01: the first version of this page named Flux Kontext at 13 credits as the instruction editor. 8frame's Flux Edit tool opens on FLUX.2 [pro] edit (7 credits for an input around 1 megapixel, more for larger inputs), with Kontext Max as the flat 13-credit option. It described Smart Edit as masked region editing; Smart Edit takes a text instruction and has no mask input. It priced the upscale step at 13, which is Clarity, a diffusion-based upscaler that always enlarges 2x on 8frame; the upscale tool opens on Topaz (from 11), and BRIA is 6. It also made quality claims we have not measured (Kontext following instructions literally, Nano Banana Pro keeping a face or product exact, dedicated background removal beating prompts "consistently") and an unsourced claim that most professionals let AI do "the first 80%". Those are removed, and the costs are recomputed. It also presented Smart Edit as a canvas tool and offered Topaz Recovery or Redefine at 65 per 24 MP: Smart Edit runs only through 8frame's MCP connector, and Recovery and Redefine are not available on 8frame (Topaz here is a fixed 4x enlargement, 11 per started 24 MP of output, up to 281).
"AI image editor" covers three different jobs and no single tool covers all three. Instruction editing ("make the jacket red, keep everything else") is one problem; object work like removing a background or replacing an element is another; finishing, meaning upscaling and cleanup, is a third. On the 8frame canvas the starting points are Flux Edit for instruction edits (FLUX.2 [pro] edit from 7 credits, or Kontext Max at 13), Nano Banana Pro at 21 when you want to rebuild the image from references, background removal at 12 for the object job, and an upscale to finish: BRIA at 6, Topaz from 11, or Clarity at 13. Full per-tool prices are in AI image editing prices by tool.
The three jobs
Instruction editing. You describe a change in words and the model applies it while leaving the rest alone. The hard part is the "leaving the rest alone": change the jacket and the face can shift too.
Object work. Removing, replacing, or isolating something. Background removal, cleanup of an unwanted element, swapping a surface.
Finishing. Upscaling, sharpening, artefact cleanup. Boring, and it is what separates output that looks generated from output that looks produced.
What each job costs
8frame credit prices, a credit is $0.01 in a top-up pack.
| Job | Tool | Credits |
|---|---|---|
| Instruction edit (tool default) | Flux Edit, FLUX.2 [pro] | 7 for an input around 1 MP; rises with input size |
| Instruction edit, flat price | Flux Edit, Kontext Max | 13 |
| Edit by recomposing from references | Nano Banana Pro, up to 10 references | 21 (42 at 4K) |
| Instruction edit on OpenAI's image model | Smart Edit (MCP connector only, not on the canvas) | 33 |
| Background removal | Background removal | 12 |
| Upscale, content-preserving, 2x or 4x | BRIA Increase Resolution | 6 |
| Upscale (tool default) | Topaz, fixed 4x, per started 24 MP of output | 11 (up to 281) |
| Upscale that adds detail, 2x | Clarity (diffusion-based) | 13 |
A full pass on one image (a FLUX.2 [pro] edit on a 1-megapixel input, background removal, a Topaz upscale) is 30 credits, about $0.30; with Kontext Max and Clarity instead it is 38. That number matters more than any ranking, because editing is iterative: the useful comparison is cost per finished image after four attempts, not per attempt.
Which to use when
Flux Edit for the common case: a described change to an existing image where the rest should stay put. The default FLUX.2 [pro] edit is billed on the input image's size plus the output, so crop before you edit; Kontext Max is the flat-price option.
Nano Banana Pro when you want to rebuild the image from your photo and other references, or change text on a label. It costs more, and as with any editor, check faces, products and lettering against the original at 100%.
Smart Edit when you work from an AI assistant through 8frame's MCP connector and want OpenAI's image model to apply the instruction; it is not on the canvas. It takes a text instruction, not a mask; 8frame's connector documentation gives its output as a 1024 x 1024 square, so expect square results from other shapes. On the canvas, GPT Image with your image attached is the way to edit with OpenAI's model.
Background removal as a dedicated step (12 credits) rather than a described instruction, so the cutout does not depend on how a prompt is interpreted.
Upscaling last, always. Flux Edit's default is priced on input size, so editing an upscaled file costs more, and edits are easier to judge and redo at generation resolution. Pick the upscaler by intent: BRIA or Topaz to enlarge, Clarity when you want added detail and accept that a diffusion upscaler can change it.
Where AI editing still fails
Text inside images. Editors can redraw words you did not ask to change. Editing a label and expecting the other words to survive is where many edits break, so check every line.
Small consistent details across a set. One image, fine. Twelve product shots that must all match after the same edit, and drift can appear.
Anything a viewer will inspect. Hands, jewellery, mechanical parts, typography. The rule from generation applies to editing: convincing at a glance is not the same as correct at full zoom.
Compositing that needs real geometry. Perspective and shadow consistency in a hard composite is usually still a job for a person in Photoshop.
The workflow that works
Edit at generation resolution, not upscaled. Do one change per pass rather than four in one prompt, because a compound instruction that half-works cannot be debugged. Keep the original, always: editing chains degrade, and going back to source beats correcting a correction.
For sets, edit one image to approval and then save the same chain as a reusable workflow to run on the rest, instead of rebuilding each edit by hand and watching them drift apart.
FAQ
What is the best AI model for editing photos? There is no single best; it depends on the job. On 8frame: Flux Edit (FLUX.2 [pro], from 7 credits) for described changes, Nano Banana Pro (21) when you work from references or need to change label text, Smart Edit (33, through the MCP connector only) for OpenAI's image model, and a dedicated tool for cutouts and upscaling.
Can AI editing replace Photoshop? For described changes and object work, often. For precise compositing, geometry, and anything requiring exact control, no. Use AI for the described changes and keep Photoshop for the precise work.
How much does an AI image edit cost? 6 to 33 credits for most operations in the table on 8frame (FLUX.2 [pro] edit rises with input size; Topaz rises with output size, up to 281), so a full edit, cutout and upscale pass is about 30 to 38 credits ($0.30 to $0.38). Budget for four attempts per finished image.
Does 8frame's Smart Edit use a mask? No. Smart Edit takes an image and a text instruction; there is no mask input.
Sources
- 8frame edit, cutout and upscale tools, defaults and prices (Flux Edit with FLUX.2 [pro] and Kontext Max, Nano Banana Pro, Smart Edit, background removal, BRIA, Topaz, Clarity): the 8frame tool registry, connector tool definitions and content fact sheet, checked 2026-10-01
Three jobs, three tools, and always upscale last.
Try it without signing up: the 8frame canvas opens on a real board. The canvas is free and unlimited; generation is paid from $19/month.