Tiled Image Watermark
Repeat a text watermark across the whole image — diagonal, staggered and impossible to crop off. Choose size, color, opacity and density. Your file never leaves your browser.
Tile a watermark across a whole image, in your browser
A watermark tucked in one corner is one crop away from gone. This tool repeats your text over the entire picture — the anti-theft pattern stock agencies put on their previews — so any crop, screenshot or re-upload still carries your mark. Pick the image, type the text, and the tiled pattern is drawn with ffmpeg locally: nothing is uploaded, and the original file never leaves your machine.
Everything about the pattern is adjustable: Text size, Text color, Opacity, the Angle of the whole grid (30° is the classic diagonal, 0 gives straight horizontal rows), how many Tiles across and Tiles down, and whether alternate rows are staggered (Brick) or aligned (Grid). Tile positions are relative, so the same settings look identical on a 400px avatar and a 6000px photo — no re-tuning per image.
Worked example
Upload beach.jpg (1600 × 1200) and use the defaults: text SAMPLE, size
32, color #ffffff, opacity 0.3, angle 30, 4 tiles across, 5 down,
Brick layout. The output is beach.jpg at the same 1600 × 1200 — forty-odd
white SAMPLE tiles running diagonally corner to corner, offset row to row,
with no bare triangles in the corners. Over a solid blue area (RGB 0, 0, 255) a
glyph pixel measures RGB (77, 77, 255): exactly 30% white composited over the
photo, so the picture still reads through the mark.
Drop Opacity to 0.15 and the same pixel measures RGB (38, 38, 255) — a
whisper-light proofing mark. Raise Tiles across/down to 6 × 8 and the
pattern becomes dense enough that cloning it out costs more than licensing the
photo.
Choosing settings
- Opacity 0.15–0.25 — proofing and client previews: visible, not intrusive.
- Opacity 0.3–0.45 — the deterrent default range for public galleries.
- Opacity 0.6+ — "do not use" stamps; the mark dominates the image.
- Angle 30–45° — diagonal text is the hardest to clone out cleanly.
Angle
0gives tidy horizontal rows, better for documents and scans. - Brick vs Grid — Brick offsets alternate rows by half a tile, so the pattern looks less mechanical and leaves no clean vertical gutter to crop along. Grid keeps every row aligned, which suits document stamps.
- Text size — start at roughly 2% of the image width (about
32for a 1600px photo,40for a 2000px one) and adjust from the preview. - Black outline — turn it on when the photo has both bright sky and dark shadow; the outline keeps the mark legible on either.
Limits and edge cases
- Text: up to 120 characters, drawn from a text file rather than pasted
into the filter, so quotes, colons, commas and
%are all safe. Use\nfor a second line. - Density: 1–12 tiles across and 1–12 down. Very long text at a high tile count will overlap between tiles — shorten the text or lower the count.
- Size: text size 6–400 px; opacity 0.02–1.0; angle −90° to +90°.
- Input: PNG, JPG, WebP, GIF and BMP up to about 25 MB. Animated GIFs keep their animation when the output format is Keep; converting to PNG/JPG/WebP takes the first frame only.
- Transparency: JPG output has no alpha channel, so a transparent PNG becomes black where it was see-through — choose PNG or WebP to keep it.
- Not a security control: a visible watermark deters casual reuse; it is not encryption, and a determined editor can still paint parts of it out. Keep an un-watermarked master.
FAQ
Will the watermark survive cropping?
That is the point of tiling. A corner watermark disappears with one crop; a tiled pattern repeats across the whole frame, so every crop large enough to be useful still contains several complete tiles. Raise Tiles across/down for smaller crops.
Can I tile a logo image instead of text?
Not here — this tool tiles text. The page takes a single uploaded image, so there is nowhere to supply a second logo file. To place a logo once (with position, scale, opacity and blend mode) use the image-composite tool, or type your studio name here as the repeating mark.
Is my photo uploaded anywhere?
No. The page loads an ffmpeg build compiled to WebAssembly and runs the whole watermarking pass inside the browser tab. The image is read from disk into memory, processed, and offered back as a download — no server round trip, which is also why it works offline once the page has loaded.
Why does the watermark look darker than the opacity I asked for?
It shouldn't — and if you have seen that in other tools, this is the reason. Naively drawing semi-transparent text onto a transparent layer blends the glyphs against black first and then again during compositing, so 30% opacity lands nearer 25% and muddied. Here the text is drawn fully opaque and the whole layer's alpha is scaled once at the end, so 30% opacity composites to exactly 30%.
What size and density should I use for a 4000px photo?
Density is relative, so Tiles across/down need no change at all — 4 × 5
covers a 400px thumbnail and a 4000px master the same way. Only Text size
is in pixels: scale it with the image, around 2% of the width, so about 80
for a 4000px-wide photo.
Does the output keep my image's format and quality?
With Output format set to Keep, the file stays in its original container
(a JPG in, a JPG out) and is re-encoded once. Choose PNG for a lossless result,
WebP for a smaller lossless-ish file, or JPG for the smallest download. JPG
output is encoded at high quality (-q:v 2).
Developer & Automation Access
Run it from the terminal
Same engine as this page, headless — via the gizza CLI:
gizza tool image-watermark-tile 'url=https://example.com/input' 'text=SAMPLE' 'font_size=32' 'color=#ffffff' 'opacity=0.3' 'angle=30' 'columns=4' 'rows=5' 'pattern=brick' 'outline=true' 'format=keep'New to the CLI? Get gizza →
Open it by URL
Pre-fill and auto-run this tool with query parameters — the names match the API/CLI:
https://gizza.ai/tools/image-watermark-tile/?url=https://example.com/input&text=SAMPLE&font_size=32&color=%23ffffff&opacity=0.3&angle=30&columns=4&rows=5&pattern=brick&outline=true&format=keepMachine-readable descriptor: tool.json — title + parameters JSON Schema for agents.