NEWVectors or files. Pick a path.Start →
    Video Understanding
    6 min read
    Updated 2026-10-23

    How Much Does It Cost to Scan Every Video Ad in Our Archive for Logos and Products?

    For 10,000 thirty-second ads (5,000 minutes), list prices on 2026-10-23 put labels, logos and on-screen text at $1,600 on Google Video Intelligence, labels at $500 on AWS Rekognition, and searchable scenes plus on-screen text at $750 on Mixpeek. Here is the arithmetic, what each price buys, and how to find every ad that shows a product.

    Ethan Steininger
    By , Founder & CEO, Mixpeek
    Video Ads
    Logo Detection
    Creative Intelligence
    Pricing
    Ad Tech
    First-Party Data

    How much does it cost to scan every video ad in our archive for logos and products?



    At list prices on 2026-10-23, scanning 10,000 thirty-second ads (5,000 minutes of video) costs about $1,600 on Google Cloud Video Intelligence for labels, logo detection and on-screen text, $500 on AWS Rekognition for labels alone, and $750 on Mixpeek for searchable scenes, transcripts and on-screen text. The bill is per minute of video per feature, so the total is minutes times the features you turn on. What differs is what you get back: annotations to store and query yourself, or an index you can search.

    What does each provider charge per minute?



    We read each provider's public pricing page on 2026-10-23 and took the first paid tier.

    FeatureGoogle Video IntelligenceAWS Rekognition VideoMixpeek
    Labels or searchable scenes$0.10$0.10$0.066 (scenes, transcript and descriptions)
    Logo detection$0.15not listednot a separate line
    On-screen text$0.15not listed for stored video$0.1098
    Object tracking$0.15
    Speech transcript$0.048included in the $0.066
    Content moderation$0.10 (explicit content)$0.10
    Free allowancefirst 1,000 minutes a month, per feature


    Twelve Labs publishes plans (a free plan with 3 GB of storage, Plus at $29 a month, Pro at $99 a month) and bills usage past the plan as overage on a separate page, so we left it out of the per-minute math. See its pricing page.

    What would 10,000 video ads cost?



    Our arithmetic for one month: 10,000 ads at 30 seconds is 5,000 minutes. Google's first 1,000 minutes of each feature are free, so 4,000 minutes are billed per feature.

    Provider and featuresCost for 5,000 minutes
    Google: labels, logos and on-screen text$1,600
    Google: the same plus speech transcription$1,792
    AWS Rekognition: labels$500
    AWS Rekognition: labels and content moderation$1,000
    Mixpeek: searchable scenes, transcript and descriptions$250
    Mixpeek: the same plus on-screen text$750
    Mixpeek: the same plus faces$1,250


    Mixpeek plans start at $25 a month (Build) and storing the resulting vectors is billed separately at $1.56 per million vectors a month. Volume tiers lower Google's rates past its first tier, and every provider negotiates at enterprise volume, so treat these as list-price ceilings.

    What does each price buy?



  1. Google and AWS return annotations. Logo detection names the logos it recognizes in each shot,
  2. with timestamps and boxes. To answer "which ads show our product", you store those results and query them yourself, and a logo or product outside the model's catalogue does not appear.
  3. Mixpeek returns a searchable index. Each scene is embedded, so you can search the archive by a
  4. product photo, a logo image or a sentence such as "a person pouring the drink into a glass", and filter on the transcript and on-screen text. Results come back as scenes with start and end times. There is no fixed logo catalogue, and nothing has to be labelled first.

    How do I find every ad that shows a specific product?



    1. Put the creative in one bucket, with the campaign, brand and flight dates as object metadata. 2. Index the video once with scene search and on-screen text, so the brand name in a super or an end card is searchable as text. 3. Search by example. Use a clean product photo as the query; image-to-scene search finds the product on a shelf, in a hand or in a background. 4. Filter and group. Filter by brand or date from the metadata, then group results by ad so each creative appears once, with the timestamps where the product shows. 5. Check the misses. Run the same search with the product name as text, which catches ads where the product is spoken or written but small on screen.

    Doing it in Mixpeek



    Mixpeek reads your files where they already live, in S3, GCS or any bucket, and turns what's in them into data your software can search, classify and moderate. For an ad archive that means one collection over the bucket, with video search and on-screen text turned on, and a retriever whose feature_search stage takes either an image or a text query.

    Who it fits: brand, agency and retail media teams that need to find creative by what is in it. Who it does not: if you need a named-logo report to hand to a sponsor, a logo detection API gives you that list directly.

    Frequently Asked Questions



    How much does AI logo detection in video cost?



    Google Cloud Video Intelligence lists logo detection at $0.15 per minute of video after the first 1,000 free minutes a month, as of 2026-10-23. Scanning 5,000 minutes of ads for logos alone comes to $600.

    Does AWS Rekognition detect logos in video?



    Its stored-video price list covers label detection, shot detection and content moderation, each per minute, and does not list a logo detection feature. Brand-specific logos on AWS usually mean training a custom model with Rekognition Custom Labels, which is priced separately.

    How do I search video ads for a product without labelling data first?



    Index the ads with an embedding model that places video scenes and images in the same space, then use a photo of the product as the query. The search returns scenes that look like the photo, with timestamps, without a trained classifier for that product.

    What is the cheapest way to analyze thousands of video ads?



    Turn on only the features you will query. At list prices, scene search alone costs $0.066 a minute on Mixpeek and label detection $0.10 a minute on Google or AWS. Each added feature, such as on-screen text or logos, adds its own per-minute rate.

    Related



  5. Creative intelligence for ad tech
  6. Best ad tech AI platforms
  7. Why logo search returns the wrong brand on transparent images
  8. Twelve Labs vs AWS vs Mixpeek for contextual advertising
  9. Managed Mixpeek

    Put multimodal search to work

    Connect a bucket and Mixpeek runs the whole multimodal search pipeline for you: extraction, indexing, and search over your own objects. No models to wire up, nothing to host.

    Start with Managed
    MVS · bring your own

    Already have vectors?

    Keep your embeddings on your own cloud and run dense, sparse, and BM25 search directly on object storage. From $25/mo.

    Start with MVS

    Search your ad archive by product, logo or line of copy

    Point Mixpeek at the bucket that holds your creative. Every scene, transcript line and piece of on-screen text becomes searchable, with timestamps.

    Creative intelligenceSee pricing

    Related guides

    Video Understanding

    How Do I Find What Makes My Best-Performing Video Ads Work? Tag Every Scene, Then Compare

    Split each video ad into scenes, tag what every scene shows (the hook, the product, people, on-screen text, the call to action), join those tags to each ad's results, and compare winners with the rest inside the same campaign and objective. Then confirm the pattern with an A/B test. Covers which attributes to tag, how to avoid crediting spend instead of creative, and how to run it on hundreds of ads.

    Read guide →
    Video Understanding

    How Does Scene-Level Contextual Targeting Work for CTV?

    Scene-level contextual targeting picks a CTV ad by the scene that plays just before the ad break, read from the video's frames, audio and captions, with no data about the viewer. Each scene is classified when the stream is ingested, the categories are stored with timestamps, and at the break the signal is a lookup. How it differs from program-level targeting, what it reads, how it reaches the bid, and who sells it.

    Read guide →
    Video Understanding

    How Do I Find Specific Moments Across Thousands of Hours of Robot or Dashcam Recordings?

    Narrow by telemetry first, skip idle footage, then search what is left by meaning. Cut continuous recordings into fixed 10 to 30 second segments, embed each segment, store its timestamps next to the vehicle or robot's sensor data, and filter before you rank. For 10,000 hours, that is 1.2 million vectors and about 15 GB of index, next to 36 TB of footage.

    Read guide →