Correction, 2026-10-01: an earlier version listed Smart Edit (33 credits) as a canvas route. Smart Edit runs only through 8frame's MCP connector, from an AI assistant; on the canvas, edits go through Flux Edit or Kontext Max. It also wrote as if those edit nodes had a strength setting: they take a prompt, one image and an aspect ratio, with no strength slider, and Flux Edit is 7 credits only for an input of about 1 megapixel.
Image to image is easy to overlook, because people reach for text-to-image by habit. If you already have something close to what you want, editing it keeps what is already right, which is far more controllable than describing the whole thing again and hoping.
TL;DR
- Free tiers exist across browser tools and open-source local options
- On 8frame: Flux Edit from 7 credits (about 1 MP input; larger images cost more), Kontext Max at 13 on the canvas; Smart Edit at 33 through the MCP connector only
- In tools with a strength (denoise) slider, it decides how much of your image survives. 8frame's edit nodes have no such slider: the prompt decides
- Edit rather than regenerate when you already have most of the image
- Your source is the ceiling. A soft input produces a soft output
What the strength setting actually does
Many image-to-image tools have a control called strength, denoise, or transformation amount. It determines how much of your original survives.
- Low. Small changes: colour, light, texture, minor cleanup. The composition and subject stay put.
- Medium. Style changes while keeping the composition. The useful middle for restyling.
- High. Your image becomes a loose suggestion. You are effectively doing text-to-image with a hint.
A common mistake: setting it high because the result looks more impressive, then losing the thing that made you start from an image in the first place. Start low and increase. The right setting is the lowest one that achieves the change.
On 8frame there is no strength slider. The edit node (Flux Edit, the default, or Kontext Max) takes a prompt, one image and an aspect ratio. The prompt does the job a slider does elsewhere: name the one thing to change, and say what must stay ("keep the composition, the framing and the product exactly as they are").
What it costs
Canon 8frame prices, a credit is $0.01 at pack rate:
| Task | Route | Credits |
|---|---|---|
| Standard edit | Flux Edit (FLUX.2 pro) | 7 for about 1 MP; 20 for a 2048 x 2048 image |
| Stronger editing pass | Kontext Max | 13 |
| Complex multi-step edit | Smart Edit (MCP connector only, not on the canvas) | 33 |
| Remove the background | background removal | 12 |
| Upscale the result | Clarity image upscale | 13 |
| Generate from scratch instead | Seedream 5 | 6 |
Note that generating a new image at 6 credits is cheaper than even the smallest edit at 7. Flux Edit bills the input's megapixels plus the output's, so crop or resize before you edit a large photo. The reason to edit is not cost per operation; it is that you keep what was already right, which saves the attempts you would spend getting back to it.
Where image to image wins
Keeping a composition you like. You have the framing and the subject; you want different light. Text-to-image will give you a different framing every time.
Working from a photograph. Your real product, your real room, your real person, adjusted rather than replaced. This is the honest route for anything a customer will encounter in reality.
Iterating on an approved direction. Once a client approves an image, variations should start from it, not from the prompt that happened to produce it.
Style transfer. Your photograph rendered as an illustration: medium strength where there is a slider; on 8frame, name the style and say the composition stays.
Fixing one thing. A distracting object, a colour that clashes, a crop that is wrong.
Where it fails
Adding something that was not there, in the right place. Tools with mask inpainting handle this better than a global edit. 8frame has no mask tool, so name the exact position in the prompt and expect a few attempts.
Text. Editing existing text in an image is unreliable. If you try it, Kontext Max (13 credits) is one route on the canvas; check every letter before it ships.
Faces, under big changes. The identity can drift. For any real person, keep the change small (low strength where there is a slider; on 8frame, say in the prompt that the face stays unchanged) and compare with the original.
Precise geometry. A product's exact shape can shift. For product work, ask for one small change at a time and check against the real item.
The workflow that uses it properly
- Generate or photograph a base image. Seedream 5 at 6 credits, or a camera.
- Fix the one thing. Flux Edit, with a prompt that names only that change and what must stay.
- Iterate on that result, not on the original prompt. Each edit is 7 credits on an image of about 1 MP (20 at 2048 x 2048) and compounds toward what you want.
- Upscale at the end, 13 credits, once.
Five edits on a 1 MP image and a Clarity upscale is 48 credits, about forty-eight cents; at 2048 x 2048 the same five edits are 100, so 113 with the upscale. Either way it converges on a specific result rather than sampling the space randomly.
FAQ
What is the best free image to image tool? A browser tool for one-off work, or an open-source model locally if you want unlimited attempts and privacy.
What strength should I use? In a tool with a slider, the lowest that achieves the change: start low and increase. On 8frame there is no slider; the prompt sets how much changes.
Can I edit text inside an image? Unreliably. Kontext Max (13 credits) is one route on the canvas; check every letter.
Why does my subject change? In a tool with a slider, strength is too high: lower it. On 8frame, the prompt asked for too much: name one change and say what must stay.
Sources
- Smart Edit as an MCP-only tool (33 credits); Flux Edit and Kontext Max on the canvas: 8frame front and back tool registries and MCP connector config, checked 2026-10-01
- Edit node inputs (prompt, one image, aspect ratio; no strength setting) and Flux Edit pricing by input plus output megapixels (calculateFlux2Cost, FLUX_KONTEXT branch): 8frame front and back tool registries, read 2026-10-01
Start from what you have, change the least.
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.