Yandex Images Scraper

Type in what you are looking for and get Yandex's picture results as a clean table. Every row is one picture: where the full-size file lives, how many pixels wide and tall it is, how many bytes it weighs, a thumbnail, its title, and the web page it appears on. Filter by size, by shape and by kind of picture. Nothing to configure, no account to create — enter your search terms and run it.

Yandex is the strongest picture index for Russian, Cyrillic and CIS content, and it indexes a very different slice of the web from the search engines most tools cover. If your work depends on finding pictures that Western indexes simply do not hold, this is the shortest route to them.

What you can do with it

• Build a picture set for machine learning. Thousands of full-size addresses with exact pixel and byte sizes, so you can select by resolution before you download anything.
• Find where your photography has been re-used. Search your product or brand names and see every page that publishes a matching picture, with its domain.
• Source product shots. Search a model name and get every listing photo the index holds, ranked as shoppers see them, with the shop page each one sits on.
• Audit picture visibility. See which sites own the picture results for your category, and where your own pictures rank against them.
• Prospect for links and partners. The domain column turns a picture search into a list of sites publishing about your subject.
• Feed a design or moodboard pipeline. Filter to large landscape photographs, or to line drawings only, and get a ready-made table of candidates.

What you get

One row per picture. An abridged real example:

{
  "query": "кошка",
  "position": 1,
  "page": 1,
  "imageUrl": "https://biosfera.vet/wp-content/uploads/2017/02/081.jpg",
  "width": 730,
  "height": 486,
  "fileSizeBytes": 96185,
  "thumbnailUrl": "https://avatars.mds.yandex.net/i?id=fc43601ccabad6df317f586f0b47c2229789132e-13307699-images-thumbs&n=13",
  "thumbnailWidth": 480,
  "thumbnailHeight": 320,
  "title": "В мире животных Ветклиника БИОСФЕРА",
  "description": "Ученые рассказал, что у кошек несколько типов личности.",
  "sourcePageUrl": "https://biosfera.vet/v-mire-zhivotnyh/",
  "sourceDomain": "biosfera.vet",
  "requestedRegionId": null,
  "regionId": null,
  "regionName": null,
  "collectedAt": "2026-08-17T11:44:02.000Z"
}

Results also come as a ready-made table view and as a spreadsheet download.

Input reference

Field	Type	Default	What it does
Search terms	list of text	—	The terms to look up. One row is produced for each different picture found. Terms in Russian return the richest results, but any language works.
Images per search term	whole number	100	How many different pictures to collect for each term. Maximum 1500; in practice the source makes roughly 1,200 different pictures available for any one term.
Image size	choice	Any size	Keep only large, medium or small pictures. Large is roughly 1,200 px wide and up, medium roughly 700 px, small roughly 250 px.
Orientation	choice	Any orientation	Keep only landscape, portrait or square pictures.
Image type	choice	Any type	Keep only photographs, clipart, line drawings, or captioned poster pictures.
Market	choice	Source default	Rank results the way people in one market see them — Moscow, Saint Petersburg, sixteen other Russian regions, Belarus and Kazakhstan. On picture search this is a ranking preference rather than a strict filter.
Search terms at a time	whole number	3	How many terms to work through at once. Higher finishes a long list sooner; lower is gentler on the source. Maximum 10.

Output fields

Field	What it is
query	The search term this picture was found for.
position	Where the source ranked it, counted from 1. Gaps appear where a repeat was removed.
page	Which result page it came from.
imageUrl	Address of the full-size picture.
width / height	Its size in pixels.
fileSizeBytes	Its weight in bytes.
thumbnailUrl	A small preview version.
thumbnailWidth / thumbnailHeight	The preview's size in pixels.
title	The title the source shows for the result.
description	The short text the source shows underneath it.
sourcePageUrl	The web page the picture appears on.
sourceDomain	That page's domain.
requestedRegionId	The market you chose, or empty if you chose none.
regionId	The market this row belongs to: the one you chose, or — when you chose none — the one the source picked by itself.
regionName	The English name of that market. Filled when you chose a market; empty when you did not, because the market the source picks by itself has no name to show.
collectedAt	When the row was collected.

Any field the source does not publish for a given picture comes back empty. It is never filled with a zero, a guess, or a value taken from another row.

Pricing

You pay per picture collected — nothing else.

What you pay for	Price
Each picture collected (one row)	$0.30 per 1,000

Worked example. Ten search terms, 100 pictures each, is 1,000 rows and costs $0.30. Taking one broad term all the way to the bottom of what the source publishes — around 1,200 different pictures — costs about $0.36.

A search term that genuinely returns nothing costs you nothing, because nothing was collected. Repeats the source shows more than once are reported once and charged once.

Limits & what this actor cannot do

• The source publishes a limited number of results for any one search term. It stops at roughly 1,200 different pictures, and it repeats some of them along the way. Repeats are reported once, so a term taken to full depth typically yields between 1,000 and 1,300 rows, not more. Asking for more than that returns everything the source makes available.
• Results are a snapshot at the moment of collection. The source's own picture results change continuously, and the same term run tomorrow will not return an identical list.
• No licence, copyright, author or capture date is available. The source does not publish rights information, photographer names, upload dates or camera details for picture results, so this actor cannot report them. Check the rights on each picture at its source page before you use it.
• Addresses point at other people's sites. A picture can be moved or taken down by whoever publishes it, so an address collected today may stop working later. The actor reports what the source publishes at the time of the run and does not download the picture files themselves.
• This is a text search, not a picture-matching search. You search with words. You cannot upload a picture and ask what it looks like.
• A term the source cannot match closely is answered with loosely related pictures rather than with nothing. Nonsense and very unusual terms still come back full. The source does not reliably say when it has widened a search, so a term returning results is not by itself proof that the results are a close match. Check a sample before you rely on an unusual term.
• The market setting nudges ranking rather than filtering. Choosing a market changes which pictures rank highest and does genuinely change the result set, but pictures from elsewhere still appear. The source does not report back which market it actually used on picture search, so this actor cannot tell you when it has ignored your choice, and it does not pretend to: regionId is the market you chose. If you pick no market, the source picks one and the row records that one — and in practice regionName is then empty on every row, because the market the source picks by itself is not one of the markets offered here and no English name is known for it. Pick a market if you want that column filled.
• A search that stops before the source runs out is reported as incomplete, with the pictures it did collect kept, rather than being presented as the whole of what the source holds.
• Square is approximate. Landscape and portrait are applied strictly by the source; square includes pictures that are close to square rather than exactly square.
• Only what the public can see. The actor does not sign in, sees nothing behind a login, and cannot reveal anything the source keeps private.
• A search the source refuses to complete is reported as a failed search, never as "no results found". The run summary counts searches that produced pictures, searches that genuinely found nothing, and searches that could not be completed as three separate figures.
• Incomplete results are detected and re-read automatically, so a result page the source hands back empty by mistake does not silently cut a collection short.
• Speed depends on the size of the job and on the source's own response times. No fixed speed is promised.
• The source's terms govern automated access. You are responsible for using the data lawfully and in line with those terms, and for handling any personal data in line with applicable privacy law, including GDPR and Russian personal data law where they apply.

FAQ

Do I need an account on the source site? No. There is nothing to sign up for and nothing to connect.

Does it need my login or password? No. It never asks for one and could not use one.

Can I schedule it? Yes. Run it on any schedule and each run produces a fresh snapshot you can compare against the last.

Why did I get fewer pictures than I asked for? Because the source ran out. It publishes a limited number of different pictures for any one term, and it repeats some of them across its own result pages. Repeats are reported once, so the honest count is lower than the raw number of results the source hands over.

Are the picture files themselves downloaded? No. You get the address of each full-size picture, its dimensions and its weight in bytes, so you can decide what to fetch and fetch only that.

Does it tell me whether a picture is free to use? No, and no picture search on this source can. Rights information is not published alongside the results. Use the source page address in each row to check the terms yourself.

Do the filters actually change the results? Yes. Every filter offered here was verified to change the result set: size shifts the typical width from roughly 250 px to roughly 1,200 px, orientation applies to every row returned, and each picture type returns a substantially different set from the same term. Filters that did nothing were left out rather than offered.