Best Video Search API Providers in 2026 (Published Prices Compared)

Video search is the category where a price list lies to you twice. The first lie is the unit: one vendor bills a request, the next bills a record, a third bills credits that it never converts into requests for you, and the same vendor can use two different units on two products sold side by side. The second lie is the surface: the nearest thing one well-known vendor sells is not video search at all but a short-form video endpoint, and its rate card sits in the same table as everyone else's as though the two returned the same thing. This post prices six options in the same unit, says what each product actually searches before putting a number beside it, and shows the arithmetic. Checked 2026-09-16.

I read every rate below off the vendor's own live pricing page on 16 September 2026 and linked each one inline. I ran no benchmark and I make no claim about anyone's speed or success rate — this is a pricing, unit and field-coverage comparison built from published documentation. Serpent is the API this site sells, and it appears here as one row among several.

How I compared these video search APIs

The job is fixed: run a video search query and read back a ranked list of videos. Every rule below exists because breaking it is how the published comparisons in this category go wrong.

• Entry rate and lowest published rate are separate columns. Quoting one vendor's volume floor against another vendor's entry price is the commonest dishonesty in API pricing writing. Both appear below, and so does the monthly plan you have to be on before the floor is reachable at all.
• Say what the product searches before pricing it. A short-form video endpoint and a general video results page are different products. Putting them in one column is not a rounding error, it is a category error, and it is the single biggest trap in this particular market.
• If a vendor never says how many videos one unit returns, the cell stays empty. I would rather publish an empty cell and name the gap than invent the number that fills it. On this table that gap is almost total, and it is a finding rather than an omission.
• A derived rate names the plan price it came from. Where a vendor prints credits rather than dollars per request, the division is shown in full underneath the table so you can redo it.
• Only published rates are claimed. Three of the vendors here also list a contact-sales tier with no published price at all — SerpApi, Bright Data and HasData. A rate nobody outside the company can read is a rate nobody outside the company can bound, so none of them gets a column, and no sentence in this post claims to beat a price that could be negotiated.

The video search API price table: dollars per 1,000 requests

Same job, same unit, every figure read off the vendor's own page. Where a vendor prices credits rather than requests, the conversion is spelled out underneath.

The arithmetic, so you can check it

• Serpent, $0.10 falling to $0.07. Video search is a flat category rate per API call — $0.10 per 1,000 on Default, $0.09 on Growth and $0.07 on Scale — and the engine you pass never changes it. Growth unlocks at a single $100 deposit and Scale at a single $500 deposit, both spendable as balance rather than burned as a subscription, so there is no monthly floor to clear before the lower rate applies. The full rate card is published rather than quoted, and the cost calculator does the multiplication for a given monthly volume.
• Scrapingdog, $1.00 at LITE and $0.1364 at the top. The pricing page sells credits, not requests, and lists the credit cost of every endpoint separately: Google Videos is 5 credits a request. Scrapingdog's LITE plan is $40 for 200,000 credits, which is $0.0002 a credit, so a five-credit video request costs $0.001 — that is $1.00 per 1,000 requests. Scrapingdog's top NOVA PRO plan is $30,000 a month for 1.1 billion credits, which works out at $0.1364 per 1,000 five-credit requests. Note what the credit ceiling means in practice: LITE's 200,000 credits are only 40,000 video searches a month, not 200,000.
• Bright Data, $1.50 per 1,000 pay as you go and $1.30 per 1,000 above the included volume. Bright Data's Google Videos card prices requests with no monthly minimum, and the $499-a-month Scale plan includes 380,000 requests and charges $1.30 per 1,000 for anything past them. That $1.30 is the price of the overage, not the price of the included volume, and quoting it as though it were the plan's blended rate is the standard sleight of hand in this category.
• SearchApi, $4.00 per 1,000 down to $1.00 per 1,000. SearchApi states its per-1,000 rate on every plan rather than leaving you to divide: $4 per 1,000 searches on the $40-a-month Developer plan, and $1 per 1,000 on Octo 5M at $5,000 a month. One caution if you build a spreadsheet from that page. The Production plan on SearchApi states $3 per 1,000, a rounded figure; the full arithmetic is $100 for 35,000 searches, or $2.8571 per 1,000. Use one or the other consistently, never both.
• SerpApi, $25.00 per 1,000 down to $1.9639 per 1,000. SerpApi publishes one ladder and no per-engine surcharge, so a Google Videos search is billed the same as a Google web search: $25.00 per 1,000 on the $25-a-month Starter plan, falling to $1.9639 per 1,000 on the Cloud 54M plan at $106,050 a month. Every engine page also carries the same line about cached searches being free and not counted, which is worth knowing if your workload repeats queries.
• HasData Short Videos, $2.46 per 1,000 down to $0.69 per 1,000 — and it is a different surface. HasData's Google Short Videos API is published at $2.46 per 1,000 requests on the $49-a-month Startup plan, $0.99 per 1,000 on the $99-a-month Basic plan and $0.69 per 1,000 on the $208-a-month Growth plan. The vendor prints those dollar figures itself, and they reconcile against its credit plans: $49 divided by 200,000 credits is $0.000245 a credit, and a ten-credit request at that price is a fraction under the printed $2.46. Read the surface label before you read the price.

Prices checked 2026-09-16, each read off the vendor's own live pricing page on that date. Scrapingdog and Bright Data figures are per request; SerpApi and SearchApi figures are per search; HasData figures are per request on its short-form video product. The Scrapingdog rates are our own division of a published plan price by a published credit volume at the published credit cost per request, not figures the vendor prints; every other figure in the table is printed by the vendor. Nothing here is a benchmark, a measured success rate or a latency claim.

Two products, one vendor, two different units

This is the finding that made the post worth writing, and it hides in plain sight on one vendor's own site. Bright Data sells a Google Videos Scraper API and a Google Images Scraper API as neighbouring products under the same SERP API family, at the same headline prices. Their unit is not the same.

• The Google Videos card is written in requests throughout: a free tier counted in requests, a pay-as-you-go rate per 1,000 requests, an included monthly volume counted in requests, and an overage rate per 1,000 additional requests.
• The Google Images card beside it is written in records: a free tier counted in records, a rate per 1,000 records, an included volume counted in records, and an overage per 1,000 additional records — the same vocabulary the vendor's site-by-site Web Scraper API uses, not the vocabulary of the SERP API family it is filed under.

Bright Data carries the same $1.50 per 1,000 pay-as-you-go headline on both cards and the same $1.30 per 1,000 above the included volume, so a spreadsheet built from the two headlines will show them as identical products at an identical price. They are not. A request is one search you issue. A record is one row the vendor hands back, and the Images card never states how many records a single image search yields — so there is no published arithmetic that turns "$1.50 per 1,000 records" into a cost per image search. On the video product the unit is the request, so the figure means what you expect it to mean. On the image product it does not, which is why image search pricing has to be reasoned about separately from video.

The practical rule: before you compare two per-1,000 prices, check that the two "1,000"s count the same thing. They frequently do not, even within one vendor, even on adjacent pages, and no comparison table in this market that I could find flags it.

What each product actually searches

The second trap has nothing to do with money. The products below search three different things — a cross-engine video results page, a short-form video surface and YouTube itself — and only the label tells you which. Nobody in this category is hiding it; nobody is highlighting it either.

Checked 2026-09-16. Each row was read from the vendor's own catalogue or documentation page on that date. The Serpent delivery figures are the ones published in our own API documentation; they are a ceiling-and-measurement pair rather than a guarantee, and the same caveat is spelled out in the video search API tutorial.

Two of those rows are worth pausing on. HasData's is a short-form product. If your job is to track what wins on Google's Shorts-style surface, it is the right tool and the price is competitive. If your job is share of voice on the ordinary video results page, it is answering a different question, and no amount of price comparison makes that go away. DataForSEO's is a YouTube product. Its YouTube SERP API also sells Comments, Video Info and Subtitles as separate engine types, which is a genuinely deeper YouTube-native product than a video results page — it is simply not the same purchase, and it belongs beside a YouTube API rather than beside a video search API. Both are useful. Neither is a substitute.

What one real month costs

Per-1,000 rates flatter whoever has the largest minimum plan. Here is the same job priced as a bill: 10,000 video searches in one month, on the cheapest published plan that covers it.

Prices checked 2026-09-16. Arithmetic: Serpent, 10,000 calls at the published Default rate of $0.10 per 1,000; Bright Data, 10,000 requests at its published pay-as-you-go rate of $1.50 per 1,000, with its separate free tier not netted off because the vendor does not state that the two stack; Scrapingdog, 10,000 requests at 5 credits each is 50,000 credits, inside LITE's 200,000; HasData, 10,000 requests at 10 credits each is 100,000 credits, inside Startup's 200,000; SearchApi and SerpApi, the smallest published plan whose search allowance covers 10,000. Plan prices are the vendors' own. The same exercise across web-search vendors is in the SERP API pricing comparison.

The ordering changes completely between the two tables, and that is the honest summary of this market. On a per-1,000 basis Scrapingdog's top plan is the cheapest third-party rate on this table. On a real 10,000-search month you cannot buy that rate at all, because it lives on a plan whose credit ceiling is more than twenty thousand times the size of the job. What you would actually pay Scrapingdog that month is $40, its LITE plan price, which puts it mid-table here rather than at the bottom. Neither number is wrong. They answer different questions, and a comparison that prints only one of them is selling you something. The costs that never appear on either table — calls that bill but return nothing, credits that expire, volume you buy and never use — are a separate problem covered in the hidden costs of a SERP API and in who pays for a failed request.

What you get back for the money

Price is the part of this market everybody measures. What the response actually contains is the part nobody does, and it is the difference between a video endpoint you can build on and one you have to wrap in defensive code.

The Serpent video response is worth describing in full because it is the only part of this comparison where I can state the shape from the contract rather than from a marketing page. Every row in results.videos[] carries eleven fields, in the same order, on every engine: position, title, url, duration, source, views, thumbnail, description, publisher, embedUrl and publishedTime. Three properties of that list matter more than the list itself:

• views is a number, not a display string. It comes back as 1200000, not as "1.2M views". If you have written the regex that strips the suffix and multiplies by a million, you can delete it — along with the bug it has where a locale renders the same count as "1,2 Mio. Aufrufe".
• publishedTime is a date, in YYYY-MM-DD form, or null when the result carried no date. Not "3 weeks ago", which is unsortable, un-diffable and wrong the moment you cache it.
• Every key is always present. A value the result did not carry arrives as null, never as a missing key, so one parser handles five engines and a strict decoder has exactly one shape to bind against. Normalising several providers into one schema is a genuinely tedious problem, which is why normalising SERP JSON across providers exists as its own post.

source and publisher are deliberately different fields: source is the host the video is served from, such as www.youtube.com, and publisher is the uploader's channel name. A result that names only the platform returns null in publisher, because the platform is already in source and inventing a channel name from a hostname would be a lie with a plausible shape. Working code that reads all eleven is in the Python walkthrough, and the monitoring shape built on top of them is in tracking video presence by engine.

What I cannot do is put the same paragraph next to the other five rows, and the reason is the finding in the table above: none of them publishes the field list or the result count on the pricing page you buy from. Several publish response samples in their documentation, which is not the same commitment as a contract, and none of them states how many videos a single billed unit returns. So a cost-per-video column for this market cannot be built from published sources, by me or by anyone else. If you are evaluating one of them, the test to run before you sign anything is not a latency benchmark: it is to send the same twenty queries to each candidate and count how many rows come back, and how many of those rows carry a duration, a view count and a publish date rather than a null.

Who is not on this table, and why

An exclusion is a finding, not an omission, so here is each one with its reason.

• DataForSEO publishes no general Google Videos endpoint. Its catalogue's engine list is Google, Bing, YouTube, Yahoo, Baidu, Naver and Seznam, and the video-adjacent product is the YouTube SERP API described above. It is priced in a genuinely different unit — a billing block of 20 results on its Search and Comments types, times three on Video Info and Subtitles — so folding it into a per-request column would misstate it in both directions depending on how many results you ask for.
• HasData ships no general Google Videos API. The Google Short Videos API in the table is the nearest product, and it is labelled as a short-form surface everywhere it appears here rather than quietly treated as video search.
• Traject Data / SerpWow is excluded on a harder ground: it never publishes how many credits one search costs. Every per-1,000 figure on its plans is therefore per 1,000 credits, not per 1,000 searches, and assuming the two are the same is exactly the kind of invented number this post exists to avoid. Two of its own pages also disagree about the product's entry price, so there is no single entry figure to quote either.
• Google's own YouTube Data API is not on the table because it is not for sale. It runs on a daily quota allocation, extra quota is granted by application rather than bought, and there is no rate card to put in a column. That makes it excellent and free at small scale and unbuyable at large scale — the trade-off is set out in the YouTube API guide.

Where the others beat us

A comparison that always concludes "buy ours" is worth nothing to a reader and nothing to a search engine, so here is the honest ledger of where the rest of this table is the better buy.

• Recurring free allowances. Bright Data's free tier is 5,000 requests a month, HasData's is 1,000 credits a month and SerpApi's is 250 searches a month. Ours is a starting allowance shared across every endpoint, not a monthly refill. If your workload genuinely is a few thousand video searches a month and never grows, a recurring free tier beats any paid rate, including this one, and you should take it.
• Buying one vendor for many surfaces. If you are already paying Scrapingdog's or HasData's monthly plan for other endpoints, video search draws from credits you have already bought. Marginal cost inside a plan you were buying anyway is close to zero, and no per-1,000 table can see that.
• Short-form video. We do not sell a dedicated short-form video endpoint. HasData does, and if that surface is your question its product is the direct answer and ours is an approximation.
• YouTube-native depth. Comments, video statistics, channel objects, playlists and subtitles are YouTube-API work, not video-search work. DataForSEO sells subtitles and comments as first-class engine types on its YouTube product. Our own YouTube coverage is a separate product on a separate rate card, and neither of them is what /api/videos is for.
• Published result counts. We publish ours; that is a point in our favour. But the flip side is that our own published figures show a single call delivering roughly 50 to 60 videos rather than the 100 the parameter allows, and we say so. A vendor that publishes nothing has not promised more — it has simply not been measured, and you should measure it before you believe either of us.

Which option fits which job

• You want a ranked video results page across several engines, at any volume. The Serpent Video Search API: $0.10 per 1,000 calls with no plan to buy, one endpoint for Google, Yahoo, Bing, DuckDuckGo and Brave, and eleven fields per row that never change shape.
• You are already a Scrapingdog customer at serious volume. Its Google Videos endpoint bills 5 credits a request, and on its top plan that is the cheapest published third-party rate on this page. It costs you nothing extra to add if you are already on a plan whose credits you are not exhausting.
• You want to start today with no subscription and no commitment. Bright Data's pay-as-you-go card has no monthly minimum, which at small volume matters more than the per-1,000 number does.
• You want one vendor for a very long list of engines and verticals. SearchApi and SerpApi both sell video alongside dozens of other surfaces on a single ladder, which is worth a premium if consolidation is the goal. SerpApi's entry price is the highest here by a wide margin, so its case is breadth and documentation rather than cost.
• Your question is really about short-form video. HasData's Google Short Videos API, and be clear with yourself that it is a different surface from the one the rest of this table prices.
• Your question is really about YouTube. A YouTube product, not a video search product. Start with the YouTube API and the guide to it, and treat video search as the complement that tells you where else a query is being answered.

FAQ

How much does a video search API cost per 1,000 requests in 2026?

On the products compared here, published third-party rates span roughly 180-fold. Scrapingdog publishes the lowest of them at $0.1364 per 1,000 requests, but that rate belongs to its top plan at $30,000 a month; its entry LITE plan is $1.00 per 1,000, on $40 a month. SerpApi's entry Starter plan is the dearest entry rate at $25.00 per 1,000 searches, because Starter is $25 a month for 1,000 searches. Serpent's own video rate card is $0.10 per 1,000 calls on Default, $0.09 on Growth and $0.07 on Scale, with no monthly plan behind any of them. Prices checked 2026-09-16.

Why do Bright Data's Google Videos and Google Images products use different units?

The vendor does not explain it, and the difference is printed on its own two product cards: the Google Videos Scraper API card is written in requests, while the Google Images Scraper API card beside it is written in records. Bright Data carries the same $1.50 per 1,000 pay-as-you-go headline on both and the same $1.30 per 1,000 beyond the volume included in the $499-a-month Scale plan, so the two headline numbers look identical. They are not comparable, because the Images card never states how many records one image search produces, while the Videos card bills the request itself. Prices checked 2026-09-16.

Is HasData's Google Short Videos API the same thing as a video search API?

No, and it is the single most important thing to know before you compare its price with anyone else's. The catalogue carries no general Google Videos product at all; what it ships is a Google Short Videos API, which searches Google's short-form video surface rather than video search as a whole. HasData's Google Short Videos API is published at $2.46 per 1,000 requests on the $49-a-month Startup plan, $0.99 per 1,000 on the $99-a-month Basic plan and $0.69 per 1,000 on the $208-a-month Growth plan. If short-form video is the thing you want, that is a real product at a real price; if you want the general video results page, it is not the product you are buying. Prices checked 2026-09-16.

Which video search API has the lowest published rate?

Among the third-party products on this table, Scrapingdog's Google Videos API at $0.1364 per 1,000 requests, and that rate is reachable only on its top plan at $30,000 a month. On the plans an individual developer would actually start on the ranking inverts: Scrapingdog's LITE plan at $40 a month is $1.00 per 1,000, SearchApi's Developer plan at $40 a month is $4.00 per 1,000 and SerpApi's Starter plan is $25.00 per 1,000. Serpent's video rate card is $0.10 per 1,000 calls with no monthly plan behind it, falling to $0.07 on Scale. Published rates only: several vendors run contact-sales tiers whose floor nobody outside the company can bound, so nothing here covers a negotiated price. Prices checked 2026-09-16.

How many videos does one video search API call return?

Almost nobody publishes the answer, which is exactly why a per-video comparison of these products cannot be built. SerpApi's Google Videos results page prints no batch size at all, even though its Google Images results page states that images are fetched in batches of 100 per search. Bright Data, Scrapingdog and SearchApi price the request or the search without saying how many videos come back inside it. Serpent publishes the measured figures in its own docs: at num=100 a single call delivered about 60 videos on google and yahoo, about 59 on bing and ddg and about 50 on brave, so num is a ceiling rather than a promise. Checked 2026-09-16.

Should I buy a video search API or a YouTube API?

They answer different questions. A video search API returns the video results a search engine shows for a query, across engines, which is what you want for share of voice, competitor monitoring and content-gap work. A YouTube API returns YouTube's own objects, such as a channel, a video's statistics, a playlist or comments, which is what you want once you already know whose channel you are measuring. DataForSEO's video-adjacent product is a YouTube SERP API rather than a general Google Videos endpoint, and it is billed for every 20 results on its Search and Comments types, which is a third unit again. If you need both, they are separate products on separate rate cards. Checked 2026-09-16.

Related reading, in the order you are likely to need it: the video search API tutorial for the first working call, the Python walkthrough for parsing the eleven fields into rows you can store, and video SERP monitoring for turning those rows into a weekly share-of-voice number.