# Image-tool comparison kit

Brainbaby · Edition 2026-10-06 · Blank planning resource · CC BY 4.0

Use the same task and instructions to compare image tools before buying more attempts. This kit contains proposed exercises and empty records, not benchmark results or a ranking. Copy it into a text editor, fill the fields, and keep it with your source and output files.

## Set the comparison before spending

- Project and intended use: [fill in]
- Review date and reviewer: [fill in]
- Tools, model labels and selected settings: [record what each tool actually displays]
- Maximum spending for this comparison, with currency: [fill in]
- Permitted source material and rights: [fill in; do not upload confidential or unauthorized files]
- Must-pass requirements: [list the details that decide whether the result is useful]
- Stop condition: [for example, stop after the agreed attempts if the product label is still wrong]

Check each tool's current upload/privacy terms, output rights, account allowance and displayed job cost. A free download of this kit does not include image jobs. Keep monetary amounts in their original currencies; do not add different currencies into one total.

## Exercise A — Generate one product concept

Use the same complete prompt in each tool:

> A blue ceramic cup on a pale studio background, soft light from the left, a subtle contact shadow, centered composition with space above for a headline. Keep the handle visible. No lettering, logo or extra objects.

Check: one cup; blue glaze; visible handle; pale background; left-side light; space above; no added text. Treat these as proposed requirements, not a promise of any tool's result. Save the exact prompt, settings and original output.

## Exercise B — Compose a scene with several constraints

> A wide editorial illustration of a small community garden: three raised beds in the foreground, a path between them, two fictional adult volunteers near the tool shed in the background. Warm late-afternoon light, clear separation between the beds, no signs or lettering.

Check: three beds; a visible path; two adult volunteers; shed in the background; wide composition; no unintended lettering. Record missing or duplicated objects instead of assigning an unsupported general quality score.

## Exercise C — Edit an authorized source photo

Use the same source file for each tool that supports editing:

> Replace only the cluttered background with a plain light-gray studio background. Keep the cup's shape, blue glaze, handle position and original crop unchanged. Match the original lighting and contact shadow. Do not add lettering, decorations or extra objects.

Adapt “cup” to your actual subject before starting; use that exact adapted wording in every tool. Compare the result with the source at full size. If a tool does not accept source photos, mark this exercise **unsupported**, rather than silently comparing a new generated concept to an edit.

## Duplicate this record for each attempt

- Tool and displayed model/version: [fill in]
- Exercise, date and attempt number: [fill in]
- Exact prompt and settings: [fill in]
- Source filename, if editing: [local reference only; do not publish private filenames]
- Displayed cost before submitting: [amount, currency or credits]
- Observed balance before/after: [fill in, or “not available”]
- Outcome: [received / failed / refused / unsupported / interrupted]
- Output filename and actual dimensions: [fill in after opening the download]
- Requirements met: [list]
- Requirements missed or unintended changes: [list]
- One proposed correction: [fill in; keep the original instruction with it]
- Final decision: [usable as-is / needs manual correction / unsuitable / not evaluated]

An interrupted request may still complete or consume allowance. Check saved job status before repeating it. Never treat an unknown result or unknown deduction as zero cost.

## Compare completed records

| Tool | Exercise | Actual attempts | Received outputs | Usable outputs | Total cost and currency | Remaining correction |
| --- | --- | --- | --- | --- | --- | --- |
| [tool] | [A/B/C] | [observed] | [observed] | [reviewed] | [actual; unknown if unverified] | [fill in] |

If you have verified the costs and have at least one usable output, divide total cost by usable outputs **within the same currency**. If there are zero usable outputs, report “no usable output”; do not divide by zero or report a zero cost per usable result. Separate generation and editing when they use different allowances. Small samples describe these particular attempts, not universal model performance.

## Share a review without leaking source material

Share the blank kit or a redacted record. Keep source photos, private prompts, account IDs and payment receipts out of a public review. Include the date, task, settings, observed limitations and any edits you made to the kit. Leave unknowns as unknowns.

Read the [matching comparison guide](https://brainbaby.ai/blog/uncensored-ai-image-generator). For source-photo work, use the [fillable edit worksheet](https://brainbaby.ai/ai-image-editing). When your brief is ready, [open Image Studio](https://brainbaby.ai/signup?redirect=%2Fimages); for editing, [open the photo-upload workflow](https://brainbaby.ai/signup?redirect=%2Fimages%3Fmode%3Dedit). Check the displayed allowance and cost before submitting. Opening a workflow does not start a paid job.

Attribution: “Adapted from Brainbaby's image-tool comparison kit, CC BY 4.0.” [Library and license](https://github.com/Brainbaby-OS/free-creative-resources).