Short answer
Choose a small, distinctive image and limit the search to the part of the screen where it belongs. Test once with the image visible and once with it absent.
Image detection reads visible pixels. It does not understand what a button means or prove that the app behind it is ready.
When to use this method
Visual conditions help when coordinates alone are unreliable. They remain sensitive to display scaling, theme, animation, layout changes and the selected region.
How we checked this guideSee what we reviewed, what we confirmed, and where the feature stops.
- What we reviewed
- Checked image and color targets, search sources, matching modes, multiple-target rules, wait modes and THEN or ELSE branches in the application source and automated tests.
- What we confirmed
- A workflow can use one or more image or color targets, wait for appearance or disappearance and run a chosen action only from a fresh verified match.
- Important limit
- When a button changes size, theme or animation state, capture a new template before replay because the saved pixels no longer describe the target.
Choose image detection or color detection
Use image detection when a recognizable shape, icon or label matters. Use color detection when one pixel color or a small group of colors is the clearest signal. An image can come from a file or the clipboard. Do not use a changing counter or clock as an image target because its visible content does not stay still.
Build a stable image condition
1. Choose a distinctive target
Use a compact part of the interface that does not include a clock, animation, cursor or changing text. Include enough shape and color to separate it from similar items nearby.
2. Choose the screen source
Search the full screen, one monitor, one visible selected window or a selected rectangle. Select a rectangle that contains the target and excludes unrelated screen areas. A selected window is still observed through its visible pixels; minimized or protected content is not a reliable source.
3. Select the match rule
Start with one expected scale and color. Use a scale range only when the same target genuinely changes size. Appearance matching emphasizes the full look, outline matching emphasizes shape and hybrid matching combines both. Grayscale can help when color is irrelevant. Confidence controls how close a result must be.
4. Decide what counts as success
With several targets, choose whether any one target is enough or all targets must be present in the same scan. If several matches exist, select the first, best, nearest or a specific numbered match. Use disappearance only when the target leaving the screen is the real signal.
5. Choose the wait and branches
A single scan checks once. A timeout checks until its time ends. Until found keeps checking, so it has no finite no-match result for ELSE. Put the intended action in THEN and use ELSE only for a finite, tested non-match. Clicking a found image uses the fresh verified position; a failed or disappearance result clears an old position instead of reusing it.
6. Test expected changes
Test one clear match and one clear non-match. Then repeat after a realistic scale, theme or window-size change. Tighten the search area before lowering confidence, because a large loose search can find the wrong object.
Checkpoint: prove match and non-match
Keep the action harmless. Run one scan with the target visible and another with it absent. Continue only if THEN and the timed no-match path behave differently and no old coordinates are reused.
Input Variation never changes which target is selected or where an Image Detection match is clicked. For this step, it can only make small timing changes within the limits you set. It cannot bypass bot protection.
What can break a match
- Display scaling or a different monitor
- Dark mode, hover states, animation or transparent edges
- A search region that is too large or no longer aligned
- Several similar targets without a deliberate match selector
Detection checks
- Test one expected match and one expected non-match.
- Confirm the action targets the intended window.
- Test a timeout or ELSE path without clicking stale coordinates.
- Keep F12 available as the permanent Emergency Stop.
Osenpa Auto Clicker
Combine editable input steps with multi-scale image matching, color targets, THEN or ELSE branches, loops and supervised scheduling.