Image Filter Effects applies six auditable color transformations — Grayscale, Sepia, Invert, Warm, Cool, and Original — to a single JPG, PNG, or WebP file in your current browser tab, with no upload and no server queue. The tool works one image at a time: you choose a file up to 25 MiB, pick an effect, set an intensity from 0% to 100%, and apply it to produce a same-size PNG whose alpha channel is preserved. Because every transformation runs locally against an in-memory canvas, processing multiple images is a matter of repeating the same three-step workflow file after file; nothing leaves your computer, there is no batch queue, and replacing a source file or changing settings invalidates the older result so you cannot accidentally download a stale preview. Each effect follows a documented formula — Grayscale uses the W3C's 0.2126, 0.7152, and 0.0722 channel weighting, Sepia uses the W3C 3×3 color matrix, Invert replaces each channel with its complement, and Warm and Cool are explicitly disclosed creative shifts rather than measured color-temperature corrections. That combination — small effect set, exposed math, preserved transparency, and local execution — is what makes the tool usable for a repeatable filter pass across many images without a hosted batch service.
How to Apply a Filter to One Image
- Choose a JPG, PNG, or WebP up to 25 MiB from your device, then select one effect — Grayscale, Sepia, Invert, Warm, Cool, or Original — from the effect list.
- Set intensity from 0% (original channels, no transformation) to 100% (full transformation), then choose Apply filter to start the local pixel pass.
- Inspect the same-size PNG preview shown on the page, confirm the important colors and the preserved alpha, then download the result; the download name reflects the chosen effect.
Filtering Multiple Images in One Sitting
The page itself processes one file at a time, so the practical answer to "bulk" is a tight, repeatable loop rather than a single button that handles a folder. In practice this is a fast workflow: open Image Filter Effects, drop in the first image, pick the effect and intensity you want to standardize across the batch, hit Apply filter, download, then drag in the next image. Because processing happens locally, the only real ceiling is your browser tab's memory, which is bounded by the documented 16-megapixel decoded limit and the 12,000-pixel edge cap; you are not waiting on a queue, an upload, or a server.
For a consistent look across many files, fix the effect and intensity before you start, then keep them locked while you swap images. The Original preset is a useful sanity check — it leaves the RGB channels unchanged and only exercises the decode and PNG export path, which confirms that a particular file decodes cleanly before you commit a real filter to it. Because each filter pass produces a same-size PNG with preserved alpha, transparent overlays, watermarks, and cut-outs survive the round trip, so you can filter whole folders of layered artwork without flattening them. The same one-image-at-a-time approach is how a number of other local-browser utilities handle multiple files, including the workflow described for watermarking many images locally, and the rhythm is similar: pick a tool, lock your settings, process, repeat.
What Each Filter Actually Changes
Every effect listed on the page is a defined color transform, not an opaque preset. The table below summarizes what each filter does to the RGB channels, how intensity changes the result, and the most important caveat for each one. Warm and Cool in particular are clearly labeled on the page as creative channel shifts, not Kelvin conversions, white-balance corrections, or ICC profile operations.
| Effect | Channel math | Intensity behavior | Key caveat |
|---|---|---|---|
| Grayscale | Weighted sum using 0.2126 R + 0.7152 G + 0.0722 B (W3C Filter Effects) | Linearly blends each source channel toward the weighted grayscale target | Green contributes more than red or blue to the resulting gray |
| Sepia | Applies the W3C 3×3 sepia color matrix to RGB | Linearly blends toward the full matrix result | Stays in the sepia hue family, not a calibrated film stock |
| Invert | Replaces each channel with 255 minus the source channel at full strength | Linearly blends toward the full inversion | Useful for negatives and mask inspection, not for archival restoration |
| Warm | Adds up to 30 red levels, adds up to 10 green levels, subtracts up to 20 blue levels | Linearly blends toward the warm target | Creative shift, not a Kelvin or white-balance calculation |
| Cool | Subtracts up to 20 red levels, adds up to 5 green levels, adds up to 30 blue levels | Linearly blends toward the cool target | Creative shift, not a Kelvin or white-balance calculation |
| Original | Leaves RGB channels unchanged | Not applicable | Useful as a decode and PNG-export sanity check |
One worked example: Grayscale at 50% intensity
Take a pure red pixel, source R = 255, G = 0, B = 0. At 100% intensity the W3C formula gives 0.2126 × 255 + 0.7152 × 0 + 0.0722 × 0 = 54.213, which rounds and clamps to 54. At 50% intensity the implementation computes the matrix target as a floating-point value first, then blends: 0.5 × 54.213 + 0.5 × 255 = 27.1065 + 127.5 = 154.6065, which rounds once to 155. The contract specifically calls this out: the full-strength target is not rounded before the partial-intensity blend, so intermediate strengths avoid the avoidable one-level channel drift you would see if both ends were rounded first. For a pure grayscale pass without an intensity slider, the dedicated Black and White Photo Converter applies the same W3C weighting.
Limits and Constraints for Batch Work
Three hard limits protect the browser tab while you work through a batch: the source file cannot exceed 25 MiB before decoding, the decoded image cannot exceed 16 megapixels, and either edge cannot exceed 12,000 pixels. Decoded pixel buffers are far larger than compressed files, so a 25 MiB JPG can still be safely handled while a much smaller but extremely wide PNG could be rejected by the edge limit. Other constraints to keep in mind when you plan a bulk pass:
| Constraint | Value or behavior |
|---|---|
| Accepted input formats | JPG, PNG, WebP |
| Maximum file size before decoding | 25 MiB |
| Maximum decoded pixel count | 16 megapixels |
| Maximum edge length | 12,000 pixels |
| Output format | Same-size PNG |
| Alpha channel | Preserved for every effect, including transparent pixels |
| Processing location | Current browser tab, in-memory canvas, no upload |
| Operations not performed | Resize, crop, rotate, sharpen, blur, denoise, face detection, subject isolation, automatic filter choice |
Two follow-on details matter when you batch. First, processing only starts when you choose Apply filter, and a zero-delay task lets the busy state render before the pixel loop begins, so the UI stays responsive even on a long pass. Second, replacing the source file or changing settings invalidates any older export and releases the temporary Object URL, so you cannot accidentally save a stale preview under a new filename. If the browser cannot decode a file, the page produces a clear error rather than a blank result, which is itself a useful signal when you are sweeping a folder of mixed sources.
When a Filter Is the Wrong Tool
Filters can quietly change readability, skin tones, brand colors, and the meaning of scientific or documentary images, so the right move is to review the full result and not rely on the preset name alone. A grayscale image does not automatically have good text contrast; a sepia treatment does not authenticate age; warm and cool shifts are not correction profiles. For color-critical print, medical, scientific, archival, or professional photographic work, the page explicitly recommends a color-managed application with the required profiles and calibration rather than this local pixel transformer. The same applies when you need to combine filtering with cropping, resizing, watermarking, or format conversion at scale — in that case, the practical move is to run Image Filter Effects as one step in a pipeline, then hand the PNG to a dedicated resizer, cropper, or converter for the next step. For readers who searched specifically for a batch service that handles hundreds of files with a single upload, a hosted batch image editor is a different category of tool with different trade-offs around upload time, server limits, and result ownership; the local-browser approach here trades raw per-click throughput for the fact that nothing is uploaded and every preset is auditable.