Convert Photo to Pencil Sketch

Turn photos into pencil drawings with the classic dodge-blend technique, in black and white or colour, with adjustable line weight. Nothing uploads.

🌐 Español

Drop your files here (.jpg, .jpeg, .png, .webp)

🔒 Private by design: your files are processed locally in your browser and never uploaded to any server.

Written and maintained by Last updated

The four-step trick behind every pencil sketch filter

This effect has been passed between photo editors for decades, and it is short enough to describe in full.

Take the photo and make a grey copy of it, weighting the channels the way human vision does rather than averaging them: about thirty percent red, fifty-nine percent green, eleven percent blue. Invert that grey copy so light becomes dark. Blur the inverted copy. Then combine the blurred inverted layer with the original grey using colour dodge, a blend mode that divides rather than adds.

The reason it looks like a drawing is a single mathematical identity. Dodging any value by its exact complement gives pure white. Across a flat region, the blur barely changes anything, so the inverted layer is still almost exactly the complement of the original and the result saturates to blank paper. Near an edge, the blur has smeared values across the boundary, the two layers no longer cancel, and what is left is a dark stroke. Everything the effect draws is a place where the blur and the original disagree.

Turning a portrait into a graphite study

  1. Drop your photos onto the box or use Choose files. JPG, JPEG, PNG and WebP are accepted, several at once.
  2. Leave Sketch style on Pencil (black & white) for the classic look, or pick Color pencil.
  3. Set Detail (blur radius in px; lower gives darker, more defined lines, higher gives a softer sketch). Start at the default of 6.
  4. Leave Intensity (% strength of the sketch effect over the original photo) at 100 for a full sketch, or lower it to keep some of the photograph showing through.
  5. Click Convert Photo to Pencil Sketch and watch the bar advance once per image.
  6. Use each Download link to save the result, then Process another to clear the queue.

My own habit is to run a portrait twice, once at detail 3 and once at detail 10, and keep whichever reads better at thumbnail size. The difference between them is much larger than the two numbers suggest.

Reading the detail slider properly

Detail is a blur radius, and blur radius behaves in a way that is initially counter-intuitive: turning the number down produces more visible drawing, not less.

At a radius of 1 or 2 the blur is so tight that almost every pixel disagrees slightly with its neighbours, so the output keeps sensor noise, skin texture and fabric weave as scribble. On a clean studio shot that can look wonderfully detailed. On a phone photo taken indoors it looks like static.

At 15 or 20 the blur is wide enough that only major boundaries survive, and the picture becomes a few pale contour lines on white. That suits a simple, high-contrast subject and looks empty on a busy one.

The default of 6 sits where most photographs behave. Values outside 1 to 20 are pulled back into that range, and anything that is not a number at all falls back to 6.

Black and white against coloured pencil

Both styles share the same blend layer, computed from grey. The choice is only about what that layer is applied to.

Choose the monochrome style and the result has no colour information left in it at all, since every channel receives the same dodged value. It is the honest graphite look, and it is what most people mean by a pencil sketch.

Choose the colour style and each channel is dodged independently against the same grey blend, which means the sketch structure is identical while the original hues bleed back through the surviving lines. Reds stay red, a blue jacket stays blue, and the whites are still white because saturation to 255 happens per channel. It reads like a coloured pencil drawing rather than a tinted photograph.

Prepare the photo before you draw it

The effect responds to contrast, so what you feed it matters more than the settings. A flat, underexposed image gives a faint sketch no matter what you do with the detail number. Push contrast first with Adjust Image, then sketch.

If the subject is lost in a wide frame, Crop Images will tighten it to a standard ratio first, and a tighter crop almost always sketches better than a wide one. For a plain monochrome conversion with no drawing at all, Convert Photo to Black and White offers grayscale and sepia and stops there.

Going in the opposite direction, AI Photo Colorizer adds colour to an old black and white photograph. And if what you actually want is clean outlines to print and fill in, Turn a Photo into a Coloring Page produces exactly that, with an A4 or Letter PDF option. The rest of the set is on the image tools hub.

See it in action

Screenshot of the Convert Photo to Pencil Sketch tool with sysfenix-sample.png (466 KB) loaded, Sketch style set to Pencil (black & white), Detail (blur radius in px; lower gives darker, more defined lines, higher gives a softer sketch) set to 6
Convert Photo to Pencil Sketch mid-process: sysfenix-sample.png (466 KB) loaded, Sketch style set to Pencil (black & white), Detail (blur radius in px; lower gives darker, more defined lines, higher gives a softer sketch) set to 6.
Screenshot of the Convert Photo to Pencil Sketch result screen showing sysfenix-sample-sketch.png ready to download (274 KB, 41% smaller)
The finished result: sysfenix-sample-sketch.png ready to download (274 KB, 41% smaller). The download link is a local blob URL — the file never leaves your device.

Frequently asked questions

Why does most of my photo turn pure white?

That is the technique working correctly, not a bug. In any area with no edge, the blurred inverted layer is almost exactly the complement of the original brightness, and dodging a value by its own complement always saturates to white. Flat sky, a plain wall and an evenly lit cheek all collapse to paper. Only places where brightness changes sharply survive as lines, which is what makes the result look drawn.

What does the detail number actually change?

The radius, in pixels, of the blur applied to the inverted copy before the two layers are combined. A small radius keeps the blur hugging every edge, so lines come out dark and tight and fine texture survives. A large radius spreads the blur wider, the two layers cancel over a bigger area, and the result gets paler and looser. The useful range is narrower than it looks; most photos land somewhere between 3 and 10.

What happens at an intensity of 0?

You get your original picture back, re-encoded. Intensity is the opacity of the sketch layer over the untouched photo, so 0 means none of the sketch is mixed in at all. That makes it a useful reference point rather than a wasted setting, since running once at 0 and once at 100 shows exactly what the effect contributed. Values in between give a shaded, half-drawn look.

How is the colour version different?

The blend layer is identical. What changes is what it is dodged against. The black and white version dodges a single grey value, so every channel ends up equal and the output is neutral. The colour version dodges each of the red, green and blue channels separately against that same grey blend layer, so the original hues survive inside the drawn structure, like a coloured pencil study rather than a graphite one.

Which extension will my results have?

JPG in, JPG out. WebP in, WebP out. PNG in, PNG out. A file ending in .jpeg comes back as .jpg, because the two are the same format and one spelling is enough. The name gains a sketch suffix, so portrait.jpg becomes portrait-sketch.jpg, and the colour style adds its own marker instead, giving portrait-sketch-color.jpg.

Will transparency survive?

Only if the output format can carry it. The alpha channel of every pixel passes through the effect untouched, so a PNG with a transparent background keeps that background transparent. Save the same image as JPG and the transparency is gone, because the format has no alpha channel to put it in. Keep PNG inputs as PNG when the cut-out matters.

Can I stop a large batch part way through?

You can press Cancel, but do not expect to keep what has already finished. Cancelling stops the loop and clears the panel, so the images processed up to that point are discarded along with the rest. If you have forty photos and only need six, queue the six.

Related tools