Charis Tsevis (@tsevis) header
Charis Tsevis (@tsevis) profile picture

Charis Tsevis

@tsevis

Computational artist and designer. Algorithmic mosaics: images built from thousands of pieces. I build my own tools (Mozaix). https://nitter.gq/t.co/b9XlHAl6ET

 Paphos, Cyprus
tsevis.com
Joined February 2008
  • Tweets45.7K
  • Following1.6K
  • Followers5K
3.3KMedia
Photo by Charis Tsevis (@tsevis)Photo by Charis Tsevis (@tsevis)Photo by Charis Tsevis (@tsevis)Photo by Charis Tsevis (@tsevis)Photo by Charis Tsevis (@tsevis)Photo by Charis Tsevis (@tsevis)Photo by Charis Tsevis (@tsevis)Photo by Charis Tsevis (@tsevis)Photo by Charis Tsevis (@tsevis)Photo by Charis Tsevis (@tsevis)
African Bricks: portraits built from thousands of fragments, inspired by the brick-and-colour patchwork of townships like Soweto, where colour is a declaration, not decoration. From across the room, a face. Up close, a wall of colour. My most collected series.
Algorithmic mosaic portrait of a young African woman in a colourful headwrap, looking straight at the viewer. Her face and wrap are built from hundreds of tiles of African textile patterns in terracotta, rose, indigo and gold. "Will You?" from African Bricks by Charis Tsevis.

ALT Algorithmic mosaic portrait of a young African woman in a colourful headwrap, looking straight at the viewer. Her face and wrap are built from hundreds of tiles of African textile patterns in terracotta, rose, indigo and gold. "Will You?" from African Bricks by Charis Tsevis.

Close-up of the same mosaic around one eye: every tile is a different printed pattern, zigzags, dots, diamonds and stripes in browns, pinks and terracotta, separated by thin dark lines.

ALT Close-up of the same mosaic around one eye: every tile is a different printed pattern, zigzags, dots, diamonds and stripes in browns, pinks and terracotta, separated by thin dark lines.

2
1
192
Beyond the human eye, 2/38 · Mathematical morphology, mid-1960s Georges Matheron and Jean Serra, at the École des Mines de Paris, started from mining: how do you measure the shape of grains and pores in a section of rock? Their answer was to treat an image as a set of shapes and probe it with another small shape, the structuring element. Erosion shrinks every shape by that probe and dilation grows it. Used one after the other, they remove specks, close gaps and pull apart things that touch. On my studio photo, a threshold and two such steps turn a busy room into a few clean black shapes, in 21 milliseconds. Designers use the same operations without the name: expand and contract a selection, spread and choke in prepress. It is still the cleanest way I know to tidy a mask. tsevis.com/blog/beyond-the-h…
Two panels: a selfie of a woman and a man in a studio, with tools hanging on the wall behind them, and the same photo reduced to a few large black and white shapes by a threshold and morphology.

ALT Two panels: a selfie of a woman and a man in a studio, with tools hanging on the wall behind them, and the same photo reduced to a few large black and white shapes by a threshold and morphology.

36
MAX 2026 by @Adobe in Miami is now LIVE! adobe.com/max.html?promoid=K…
Image from Charis Tsevis (@tsevis)
2
2
181
Same footage, four decisions. The beans never change.
1
1
108
Masking tools often mean a cloud service, and your client's artwork on someone else's server. Skiagrafia segments an image into semantic layers (objects and their parts) with local models, and exports SVG and TIFF/PNG. The vision-language model is only allowed to talk to localhost. github.com/tsevis/skiagrafia
The Skiagrafia desktop interface, showing an image segmented into semantic layers and masks.

ALT The Skiagrafia desktop interface, showing an image segmented into semantic layers and masks.

55
So jealous of everyone going to MAX. Adobe and Miami is a killing combination. Enjoy, everyone!
Rufus Deuchler (@rufusd) profile pictureRufus Deuchler
@rufusdNov 9
Adobe MAX starts tomorrow in Miami Beach, November 10 to 12. Not in Miami? You can follow the keynotes online for free: max.adobe.com What are you most curious to see this year?
1
118
A picture is worth a thousand words? We turned it around. Written Faces: seven portraits made of handwriting, thousands of signatures, words and notes cut from manuscripts, petitions, scores and letters. Each has a short film. tsevis.com/portfolio/a-pictu…
Portrait of the poet C. P. Cavafy made of fragments of his handwritten Greek manuscripts

ALT Portrait of the poet C. P. Cavafy made of fragments of his handwritten Greek manuscripts

Portrait of Frederick Douglass made of thousands of handwritten signatures from Massachusetts anti-slavery petitions

ALT Portrait of Frederick Douglass made of thousands of handwritten signatures from Massachusetts anti-slavery petitions

Portrait of Ludwig van Beethoven made of handwritten words in pencil and ink from his conversation books

ALT Portrait of Ludwig van Beethoven made of handwritten words in pencil and ink from his conversation books

Portrait of Florence Nightingale made of handwritten words from her letters, in pinks and whites on a dark purple ground, with rays of words spreading from a lamp's light

ALT Portrait of Florence Nightingale made of handwritten words from her letters, in pinks and whites on a dark purple ground, with rays of words spreading from a lamp's light

2
105
Diophantine, a small app for encrypting files before you send or store them: a simple window over 7z, GPG and VeraCrypt. No cloud, no accounts. For a contract by email or a private backup. github.com/tsevis/diophantin…
Diophantine's window in dark mode: Encrypt and Decrypt tabs, a choice of ZIP, 7z, GPG or VeraCrypt, a password field with Generate, keyfile and recovery-phrase options, and a drop area for files.

ALT Diophantine's window in dark mode: Encrypt and Decrypt tabs, a choice of ZIP, 7z, GPG or VeraCrypt, a password field with Generate, keyfile and recovery-phrase options, and a drop area for files.

64
Two layers of one portrait: a loose fill, then a thin skeleton of details. Right: the result.
Three panels. Left: loose scattered app icons on grey, the fill layer of a portrait. Middle: a thin outline of eyes, brows, nose and braids in tiny icons on grey. Right: the finished portrait of a young woman with braids made of app icons, in warm reds against blue and green.

ALT Three panels. Left: loose scattered app icons on grey, the fill layer of a portrait. Middle: a thin outline of eyes, brows, nose and braids in tiny icons on grey. Right: the finished portrait of a young woman with braids made of app icons, in warm reds against blue and green.

53
Cropping 300 portraits to one consistent framing per person is a detection problem, not a design problem. I'm building a local, offline cropper for the Mac with two modes: one framing per subject across a whole set, or crop to what you describe in words, using open-vocabulary detection.
1
1
74
A hybrid mosaic keeps both ideas. The darkest classes of the camouflage stay flat paint and hold the shape together. The lighter ones become photographs, so the face reads like a two-tone poster lit from inside by pictures. Made with Mozaix.
Four stacked bands of one mosaic, from a close detail of pale photographs and black paint at the top to the whole profile portrait of Abbott Thayer at the bottom.

ALT Four stacked bands of one mosaic, from a close detail of pale photographs and black paint at the top to the whole profile portrait of Abbott Thayer at the bottom.

1
2
120
Remember KPT? 💙 Kai Krause’s little plug-ins from the 90s looked like they’d arrived from the future. We live in that future now, so I went back for a moment. Burg Vector is my take on KPT Vector Effects for Illustrator, made without a single line of his code or reverse engineering. Those who know Kai will get the name. Open source and free, for Mac and Windows. Enjoy: github.com/tsevis/burg-vecto… #Illustrator #KPT #KaiKrause
Image from Charis Tsevis (@tsevis)
Image from Charis Tsevis (@tsevis)
4
157
A mosaic as old glaze. Cracks instead of grout.
A portrait of a young woman with braids made as a mosaic of small irregular cells in warm oranges, pinks, purples and teal, separated by thin dark cracks like old ceramic glaze.

ALT A portrait of a young woman with braids made as a mosaic of small irregular cells in warm oranges, pinks, purples and teal, separated by thin dark cracks like old ceramic glaze.

61
Beyond the human eye, 1/38 · Roberts cross, 1963 In his doctoral thesis at MIT, Lawrence Roberts tried to make a computer understand a photograph of simple blocks. One step on the way was to find edges. His answer was two tiny grids of numbers, two by two. Each compares a pixel with its diagonal neighbour, and where the difference is large, there is an edge. That's the whole method, and it is one of the first edge detectors. On my studio photo it takes 12 milliseconds and draws every hard border: the tools on the wall, the frames, the table, the outlines of two people. It knows nothing about what any of them are. For about fifty years computer vision was this: people deciding what an edge is and writing it down as arithmetic. I still use these methods every day. tsevis.com/blog/beyond-the-h…
Two panels: a selfie of a woman and a man in a studio, with tools hanging on the wall behind them, and the same photo drawn as fine grey edge lines by the Roberts cross.

ALT Two panels: a selfie of a woman and a man in a studio, with tools hanging on the wall behind them, and the same photo drawn as fine grey edge lines by the Roberts cross.

1
78
In my mosaic matcher DINOv3 is an opt-in feature extractor. It runs on Apple's MPS and falls back to a torchvision ViT or CNN when the weights are missing. A question for anyone who has measured it: for tile-level similarity, where the "object" is a 40 px fragment, do DINOv3 features beat CLIP?
102
New article: Beyond the Human Eye. What computer vision can do for anyone who works with images. A machine can now look at a flat photograph and tell how far away each point is, where the figure ends and the ground begins, and how an illustrator might draw it. It can find the one picture among thousands that is about the same thing, and read the titles on a shelf of books. All of it ran on one desktop computer, offline, almost all of it with open tools, on my own photos and works. For designers, illustrators, photographers, art directors and artists: how each method works, what I made with it, and where it fails. tsevis.com/blog/beyond-the-h… A companion series starts today, Beyond the human eye: 38 posts, two a week, one method at a time.
Four panels of the same photo of a woman on a staircase hanging a large mosaic portrait of a man made of octagonal tiles: the photo; its depth map in greys, the portrait and the woman light, the far wall dark; the woman and the portrait cut out together on white; and a fine line drawing of the whole scene

ALT Four panels of the same photo of a woman on a staircase hanging a large mosaic portrait of a man made of octagonal tiles: the photo; its depth map in greys, the portrait and the woman light, the far wall dark; the woman and the portrait cut out together on white; and a fine line drawing of the whole scene

1
99
#computervision
26
Every Voice. Every Story. Every Table. My key art for the American Association of Critical-Care Nurses' 2026–27 theme: three nurses, made of several thousand photographs of real nurses at work. tsevis.com/portfolio/aacn-ev…
Mosaic key art of three smiling nurses in blue scrubs standing shoulder to shoulder, two women and a bearded man, built from small photographs of nurses, with photographs drifting away at the edges.

ALT Mosaic key art of three smiling nurses in blue scrubs standing shoulder to shoulder, two women and a bearded man, built from small photographs of nurses, with photographs drifting away at the edges.

Detail of the key art at full size: small photographs of nurses at work and at a conference form a smile.

ALT Detail of the key art at full size: small photographs of nurses at work and at a conference form a smile.

Mosaic portrait of a smiling nurse built from small healthcare icons in blue, violet and turquoise, with books, syringes, monitors and hearts fading into a pale background.

ALT Mosaic portrait of a smiling nurse built from small healthcare icons in blue, violet and turquoise, with books, syringes, monitors and hearts fading into a pale background.

Detail of the icon portrait: a smiling nurse’s eyes, nose and mouth made of small blue, violet and peach healthcare icons, with clipboards, books and hearts at the left.

ALT Detail of the icon portrait: a smiling nurse’s eyes, nose and mouth made of small blue, violet and peach healthcare icons, with clipboards, books and hearts at the left.

1
1
84
The mosaic is 5,000 years old: the temple walls of Uruk, the Greek floors at Pella, the vaults of Hagia Sophia. Then Bell Labs, 1967. I make algorithmic mosaics in that line, and I think it should be said out loud. The tradition is what makes the new work readable. tsevis.com/tradition
A pebble mosaic floor from Pella, ancient Macedonia, with pale figures against a dark ground.

ALT A pebble mosaic floor from Pella, ancient Macedonia, with pale figures against a dark ground.

69
Same camouflage grid, two fills. Top: flat paint. Bottom: about two thousand photographs of soldiers. The grid decides the shape of each piece. The photograph decides what it is made of.
Two halves of one image. Top: a portrait built from small tan, olive and brown dots. Bottom: the same portrait built from thousands of tiny photographs of soldiers.

ALT Two halves of one image. Top: a portrait built from small tan, olive and brown dots. Bottom: the same portrait built from thousands of tiny photographs of soldiers.

59
CrewListr Pro, a Mac and iPhone app for yacht charter: photograph the passports, get the crew list the port authority asks for. Read on the device, kept encrypted, checked by a person before export. github.com/tsevis/crewlister…
CrewListr Pro on the Mac with sample data: three people on a sailing yacht, two cleared for export; the selected passport's fields (name, document number, nationality, dates) each confirmed with a blue check.

ALT CrewListr Pro on the Mac with sample data: three people on a sailing yacht, two cleared for export; the selected passport's fields (name, document number, nationality, dates) each confirmed with a blue check.

CrewListr Pro's fleet screen with sample yachts: name, flag, port of registry and registration number held once per yacht, with a preview of how they print on the crew list.

ALT CrewListr Pro's fleet screen with sample yachts: name, flag, port of registry and registration number held once per yacht, with a preview of how they print on the crew list.

1
77

No more items