AI detector
A classifier that estimates how likely a text was produced by a language model, from statistical patterns of the writing. It returns a probability, not proof.
The terms you meet on the tool pages and in the reports, explained in one paragraph each.
A classifier that estimates how likely a text was produced by a language model, from statistical patterns of the writing. It returns a probability, not proof.
A style transformation applied by a model — cartoon, anime, watercolour, sketch — that changes the rendering of a picture while keeping its content recognisable.
The proportion between the width and the height of a picture — 1:1 for a square, 4:5 for a portrait post, 3:2 for a standard photo. Changing the ratio means cropping, not resizing.
Separating the subject of a picture from everything behind it. The result is a cut-out with a transparent background that can be placed on a colour or another scene.
An ID photo that meets the machine-readable requirements of a passport or ID card: fixed dimensions, head size and position, neutral expression, plain background, even lighting. Automated systems compare it with the face at the border.
The variation of sentence length and structure across a text. Human writing alternates long and short sentences; generated text often keeps an even rhythm.
A reference that lets the reader find the source of a quotation or an idea: author, title, year and location, formatted in the style required (APA, MLA, Chicago…).
Adding plausible colours to a black-and-white picture, based on what the model recognises in it. Natural for skin, sky and foliage; an interpretation for specific objects.
Removing the grain produced by low light or old sensors while keeping the real edges and textures of the picture.
The print density of an image. At 300 dpi, a 35 × 45 mm photo needs 413 × 531 pixels. A higher pixel count at the same print size gives finer detail; a low pixel count prints blurry.
A portrait framed on the head and shoulders, with controlled light and a neutral background, used for professional profiles, CVs and team pages.
Restating an idea in different words and structure. Legitimate when the source is cited; a form of plagiarism when it is not.
A measure of how predictable the words of a text are for a language model. Generated text tends to have low perplexity; human writing, with its unexpected choices, tends to be higher.
Presenting someone else's words or ideas as your own without attribution — by copying, by stitching phrases from several sources, or by paraphrasing without a citation.
The text description given to an image generator or a language model. A good prompt names the subject, the setting, the light and the style in concrete words.
The share of a text that matches passages found in other sources. It includes properly quoted material, so the highlighted passages must be read, not just the percentage.
Increasing the pixel dimensions of a picture while reconstructing plausible detail, instead of simply enlarging the existing pixels. Used to print small images or display them on large screens.
A visible or invisible mark added to a picture or a text to indicate its origin. Some generators embed invisible watermarks that detectors can read.
Before
After
Passport, ID card and visa photos at the official size, with a plain background.
Learn more
Before
After
Studio-style portraits for LinkedIn, CVs and company pages, generated from a few selfies.
Learn more
Before
After
Natural, flattering profile pictures for dating apps, generated from your own photos.
Learn more
Before
After
Retouch, crop, remove objects and adjust colours in your browser, with AI assistance.
Learn more
Before
After
Sharpen blurry pictures, reduce noise and upscale small images automatically.
Learn more
Before
After
Cut out the subject of a picture and replace the background with a colour or another scene.
Learn moreChoose a tool, upload your file or paste your text, and get a result you can download right away.