Advertising

Part of the Creative DNA solution →v1.0 · updated 2026-09-11

Creative DNA for Ad Archives

Search every scene of every ad you have run, and find the ones that keep earning spend

Deploy this namespace Book a workshop

Applies empty and reads Your own ad exports. Connect your own data before querying.

scenes across 271 source videos: 2,306
face documents from the same archive: 8,168
brands in the corpus: 9
creatives carrying ad copy, out of 271: 38
median scene length: 2.5s
click to first answer: 8.2 s

How these were measured

2,306 scenes across 271 source videos: Measured 2026-09-11 by paging the live ads-scenes collection (col_a96402bd17) to exhaustion. Median 5 scenes per video, mean 8.5, max 19; 39 of 271 videos yield exactly one scene.
8,168 face documents from the same archive: Live count on col_6cdf041caf, face_identity_extractor@v1 with scrfd_2.5g detection and quality scoring on.
9 brands in the corpus: 13 distinct brand_slug values, of which four are re-mints of the same brand under a -2 suffix, so nine brands by name.
38 creatives carrying ad copy, out of 271: Live count on ads-text (col_e682a390cd), measured 2026-09-11: 38 documents, 38 distinct ad_id, one chunk each, median 184 characters. Copy is supplied per creative rather than extracted, so coverage is whatever your own export carries. On the measured corpus that is 14%.
2.5s median scene length: At scene_detection_threshold 0.5 on source clips whose own median length is 29 seconds. Mean scene length 3.6s, max 29.5s.
8.2 s click to first answer: from clicking apply to the first search that answers, timed on production on 6 November 2026 by the harness that runs nightly

What’s inside

Extractors: multimodal text
Models: Multilingual E5 Large Instruct Vertex AI multimodalembedding@001
Retrievers: brand-scenes·feature_searchcopy-search·feature_searchcompetitor-product-monitor·feature_search → llm_filter
Marketplace retrieversHosted checks that run beside it, no key: IP screening: faces and brand marks Celebrity face match Brand logo match Trademarked slogan match
Solution: Creative DNA →

See it work

ads holding the product up to camera
brand-scenes
scene embedding over the ad archive, filtered by brand_slug
which ads open on a price claim
copy-search
the spoken and on-screen copy, searched as text

Simulated walkthrough · illustrative frames, scripted decisions

Two people filming a piece to camera with a ring light on a table

ads-scenes and ads-faces, both fed from the same objectStill: Pexels / ANTONI SHKRABA productionframe 1 of 3

Detections on this frame

Decision path

ads-scenes and ads-faces, both fed from the same object
Face indexed
Scene embedded
route: cleared

Indexed as one scene. On a dedicated deployment the face is indexed as well, and each is queryable on its own terms.

Running tally

cleared

review

excluded

What a person's call does here

None. No thumbs, no saves, no relevance marks are recorded by the template as shipped.

Try a search

Pick a search to see which retriever answers it and whether it works on a one-click deploy.

Inad-creativesOutads-scenesads-facesads-text

How it's wired

Applies 1 buckets, 2 collections and 3 retrievers today; dashed parts ship commented out

SourceBucketCollectionRetrieverClusterAlertNot applied yet

Drag nodes to rearrange. Drag the background to pan, ctrl or cmd plus scroll to zoom. Click a node for its settings.

What it costs

Files per monthMinutes per file333 hours of video a monthTokens per text fileSearches per month

Usage cost per month

$1,629

$0.16 per file · rate card when this page was built

Processing is priced the way a batch is charged: each bucket's minutes, images, pages or tokens times its rate card lines. The batch dry run in Studio gives the binding quote.

Models

Multilingual E5 Large InstructServed on MixpeekEmbedding
Vertex AI multimodalembedding@001Provider APIEmbedding

Collections and their documents

ads-scenesfrom bucket ad-creatives, by multimodal_extractor@v1

Reads video from creative_asset_url

Field	Type	Required
document_id	string	set by Mixpeek
collection_id	string	set by Mixpeek
ad_id	string	yes
brand_slug	string	yes
brand_name	string	yes
ad_type	string	no
source	string	no
source_url	string	no
headline	string	no
primary_text	string	no
published_at	string	no

mixpeek://multimodal_extractor@v1/vertex_multimodal_embedding from Vertex AI multimodalembedding@001

ads-textfrom bucket ad-creatives, by text_extractor@v1

Reads text from primary_text

Field	Type	Required
document_id	string	set by Mixpeek
collection_id	string	set by Mixpeek
ad_id	string	yes
brand_slug	string	yes
brand_name	string	yes
ad_type	string	no
source	string	no
source_url	string	no
headline	string	no
primary_text	string	no
published_at	string	no

mixpeek://text_extractor@v1

Retrievers

brand-scenes
feature_search over ads-scenes; input query
copy-search
feature_search over ads-text; input query
competitor-product-monitor
feature_search then llm_filter over ads-scenes; input query, document_id

Spend limits

Daily spend budget in the manifest: 1 USD a day; new batches pause once it is spent
Acceptance run ceiling: 1 USD
Model cost: Unmeasured for 2 of 2 models

Monthly cost at your volume is in the calculator. Served models start at zero replicas, so an idle namespace runs no model.

Reward signals

Nothing in this template writes interaction signals back, and it is worth saying so rather than implying a flywheel that is not wired.

How reviewer decisions move the thresholds

DocsInteractions API

Explicit signals

None. No thumbs, no saves, no relevance marks are recorded by the template as shipped.

Implicit signals

None from the application. Mixpeek records retriever executions server-side in mxp_retriever_executions, which is telemetry about queries rather than feedback about results.

Where they land

mxp_retriever_executions

How the loop closes

Adding a learning loop means posting to /v1/retrievers/interactions when a user opens a result. That is a deliberate next step, not something this template does for you.

An ad archive you can query by what is on screen. Every creative is cut into scenes at ingest and the ad copy is searchable beside them, so a question like 'the hook where someone holds the product to camera' returns the moment rather than the file.

One reference query returns ranked scenes with their ad, brand, start and end time, so a result is something you can open at the right second.
Face search, finding every ad one creator appears in, runs on a dedicated or customer-hosted deployment. Face recognition is off on Mixpeek Cloud, so one click there deploys scenes and copy, and the face step ships commented out in the manifest.
The ad copy lives in its own collection, so search reaches what was WRITTEN as well as what was SEEN, for the creatives whose export carries copy.
Clusters group scenes into recurring visual themes per brand, which is what turns an archive into a view of what a brand keeps making.

For: Performance marketing and creative teams sitting on an ad archive nobody can search

What goes in

ad-creatives: One object per creative: the video plus the ad metadata. ad_id, brand_slug and brand_name are required; the copy, headline, source and publish date are optional and travel with it.

What comes out

ads-scenes: Scene-level documents with a multimodal embedding, timing, and the brand and ad they came from.
ads-faces: Face identity vectors, one document per detected face, with a quality score. Dedicated or customer-hosted deployments only; it is not created on Mixpeek Cloud.
ads-text: Chunked ad copy, searchable on its own or beside the visual match. Populated only for creatives whose export carries copy.

DocsManifest schema

One file spins up the namespace. Generated from the workflow above. Also served at /templates/creative-dna.namespace.yaml.

# creative-dna: one manifest spins up the namespace.
# Platform manifest schema (GET /v1/discovery/schema). Validate with POST /v1/manifest/validate,
# apply with POST /v1/manifest/apply or the Deploy button. Wiring comes from the flow diagram:
# edges are bucket -> collection sources, collection -> retriever scope, retriever -> view.
# Applying a SECOND time, over a namespace this template already created: use
# POST /v1/manifest/apply?mode=create_missing, which creates what is missing and leaves
# what exists alone. The default, create_only, fails the WHOLE apply and rolls it back if
# any resource already exists, so an upgrade looks like a dead end without this. Use
# mode=upsert to also patch resources that exist but have drifted from this file.
version: '1.0'
metadata:
  name: creative-dna
  description: "Namespace template creative-dna. Generated from the flow diagram on mixpeek.com/templates/creative-dna."
namespaces:
  - name: creative-dna
    description: "Everything below lives in this namespace."
    feature_extractors:
      - name: multimodal_extractor
        version: v1
      - name: text_extractor
        version: v1
    # Daily spend budget, in dollars of what you are charged (1,000 credits a dollar).
    # New batches pause once the namespace has spent this much today; queries are not capped.
    # Raise it later on the namespace: PATCH /v1/namespaces/<id> {"spend_budget": {"daily_usd": N}}.
    budget:
      daily_usd: 1

# Data sources. A storage connection carries credentials, so it is created in Studio
# (or POST /v1/organizations/storage-connections) and synced into the bucket named here.
#   ad-library-pull: manual, one-shot, your ad creatives and their copy -> bucket ad-creatives
buckets:
  - name: ad-creatives
    namespace: creative-dna
    description: "Meant to be fed by ad-library-pull (manual, one-shot): upload files here or write them through the API. Nothing arrives until you do."
    schema:
      properties:
        ad_id:
          type: string
          required: true
        brand_slug:
          type: string
          required: true
        brand_name:
          type: string
        source:
          type: string
        source_url:
          type: string
        headline:
          type: string
        primary_text:
          type: string
          required: true
        ad_type:
          type: string
        creative_asset_url:
          type: video
          required: true
        published_at:
          type: string
collections:
  - name: ads-scenes
    namespace: creative-dna
    description: "multimodal_extractor@v1 over bucket ad-creatives. Feeds competitor-product-monitor, competitor-product-shown, brand-scenes, creative-patterns."
    source:
      type: bucket
      bucket: ad-creatives
    feature_extractor:
      name: multimodal_extractor
      version: v1
      input_mappings:
        video: creative_asset_url
      field_passthrough:
        - source_path: ad_id
          required: true
        - source_path: brand_slug
          required: true
        - source_path: brand_name
          required: true
        - source_path: ad_type
        - source_path: source
        - source_path: source_url
        - source_path: headline
        - source_path: primary_text
        - source_path: published_at
    alert_applications:
      - alert: competitor-product-shown
        input_mappings:
          - input_key: query
            source: {source_type: constant, value: "competitor product shown"}
          - input_key: document_id
            source: {source_type: document_field, path: _internal.document_id}
    enabled: true
  - name: ads-text
    namespace: creative-dna
    description: "text_extractor@v1 over bucket ad-creatives. Feeds copy-search."
    source:
      type: bucket
      bucket: ad-creatives
    feature_extractor:
      name: text_extractor
      version: v1
      input_mappings:
        text: primary_text
      field_passthrough:
        - source_path: ad_id
          required: true
        - source_path: brand_slug
          required: true
        - source_path: brand_name
          required: true
        - source_path: ad_type
        - source_path: source
        - source_path: source_url
        - source_path: headline
        - source_path: primary_text
        - source_path: published_at
    enabled: true

# NOT APPLIED. Face recognition is turned off on Mixpeek Cloud, so this face step does not deploy. It runs on a dedicated or customer-hosted deployment, where face_identity_extractor is served.
# The block below is correct and is emitted commented out so the rest of this
# manifest applies; an apply is all-or-nothing and would otherwise roll back.
#   - name: ads-faces
#     namespace: creative-dna
#     description: "face_identity_extractor@v1 over bucket ad-creatives. Feeds cast-identification."
#     source:
#       type: bucket
#       bucket: ad-creatives
#     feature_extractor:
#       name: face_identity_extractor
#       version: v1
#       input_mappings:
#         video: creative_asset_url
#       field_passthrough:
#         - source_path: ad_id
#           required: true
#         - source_path: brand_slug
#           required: true
#         - source_path: brand_name
#           required: true
#         - source_path: ad_type
#         - source_path: source
#         - source_path: source_url
#         - source_path: headline
#         - source_path: primary_text
#         - source_path: published_at
#     enabled: true
retrievers:
  - name: brand-scenes
    namespace: creative-dna
    description: "Searches ads-scenes across 1 feature index."
    collections:
      - ads-scenes
    input_schema:
      query:
        type: text
        required: true
        description: "What to look for; searched across every index below"
    stages:
      - stage_name: search
        stage_id: feature_search
        parameters:
          searches:
            - feature_uri: "mixpeek://multimodal_extractor@v1/vertex_multimodal_embedding"
              query:
                input_mode: text
                value: "{{INPUT.query}}"
              top_k: 100
          fusion: rrf
          final_top_k: 100
    tags:
      - template:creative-dna
  # NOT APPLIED. Face recognition is turned off on Mixpeek Cloud, so this face step does not deploy. It runs on a dedicated or customer-hosted deployment, where face_identity_extractor is served.
  # - name: cast-identification
  #   namespace: creative-dna
  #   description: "Searches ads-faces across 1 feature index."
  #   collections:
  #     - ads-faces
  #   input_schema:
  #     query:
  #       type: text
  #       required: true
  #       description: "URL of the image or frame to match; encoded by the index below and compared against it"
  #   stages:
  #     - stage_name: search
  #       stage_id: feature_search
  #       parameters:
  #         searches:
  #           - feature_uri: "mixpeek://face_identity_extractor@v1/insightface__arcface"
  #             query:
  #               input_mode: content
  #               value: "{{INPUT.query}}"
  #             top_k: 50
  #         fusion: rrf
  #         final_top_k: 50
  #   tags:
  #     - template:creative-dna
  - name: copy-search
    namespace: creative-dna
    description: "Searches ads-text across 1 feature index."
    collections:
      - ads-text
    input_schema:
      query:
        type: text
        required: true
        description: "What to look for; searched across every index below"
    stages:
      - stage_name: search
        stage_id: feature_search
        parameters:
          searches:
            - feature_uri: "mixpeek://text_extractor@v1"
              query:
                input_mode: text
                value: "{{INPUT.query}}"
              top_k: 50
          fusion: rrf
          final_top_k: 50
    tags:
      - template:creative-dna
  - name: competitor-product-monitor
    namespace: creative-dna
    description: "The alert's check: does the ad scene just written show a product from a brand other than the advertiser's. The search is scoped to that one scene, and an llm_filter watches it and keeps it only when it does. Materialises into competitor-product-shown."
    collections:
      - ads-scenes
    input_schema:
      query:
        type: text
        required: true
        description: "What to look for; searched across every index below"
      document_id:
        type: text
        required: true
        description: "The written document this run is scoped to, supplied by the alert attachment"
    stages:
      - stage_name: search
        stage_id: feature_search
        pre_filters:
          AND:
            - field: _internal.document_id
              operator: eq
              value: "{{INPUT.document_id}}"
        parameters:
          searches:
            - feature_uri: "mixpeek://multimodal_extractor@v1/vertex_multimodal_embedding"
              query:
                input_mode: text
                value: "{{INPUT.query}}"
              top_k: 5
          fusion: rrf
          final_top_k: 5
      - stage_name: judge
        stage_id: llm_filter
        parameters:
          criteria: "Keep this ad scene only if it clearly shows a product or its packaging from a brand other than {{DOC.brand_name}}, the advertiser. Discard scenes that show no product, or only {{DOC.brand_name}} products."
          include_media: true
    tags:
      - template:creative-dna
clusters:
  - name: creative-patterns
    namespace: creative-dna
    description: "Groups ad scenes across brands by what they show, so recurring visual patterns can be browsed and named."
    source_collections:
      - ads-scenes
    cluster_type: vector
    vector_config:
      feature_uris:
        - mixpeek://multimodal_extractor@v1/vertex_multimodal_embedding
      clustering_method: kmeans
      algorithm_params:
        n_clusters: 6
    llm_labeling:
      enabled: true
      provider: google
      model_name: gemini-2.5-flash
      labeling_inputs:
        input_mappings:
          - {input_key: text, source_type: payload, path: headline}
          - {input_key: image_url, source_type: payload, path: thumbnail_url}
alerts:
  - name: competitor-product-shown
    namespace: creative-dna
    description: "Emails the org notifications address (Settings -> Notifications; unset means opted out) when a newly indexed ad scene shows a competitor's product."
    source: retriever
    retriever: competitor-product-monitor
    trigger_on: results
    notification_config:
      channels:
        - channel_type: email
      include_matches: true
      include_scores: true
    enabled: false

# Eight public-domain television commercials from the Prelinger Archives on
# archive.org, ingested by one bounded batch after apply, so a first deploy
# can search real ads. Each seed object carries the bucket's required fields
# beside its video. Two brands share a product in three pairs (coffee, cereal,
# soap and detergent), which is what the competitor and copy searches compare.
seed:
  - bucket: ad-creatives
    max_objects: 8
    objects:
      - blobs: {creative_asset_url: "https://archive.org/download/NorelcoSpeed/NorelcoSpeed_512kb.mp4"}
        ad_id: sample-norelco-speedshaver
        brand_slug: norelco
        brand_name: "Norelco"
        source: archive.org
        source_url: "https://archive.org/details/NorelcoSpeed"
        headline: "Norelco Speedshaver"
        primary_text: "A vintage television commercial for the Norelco Speedshaver electric shaver."
        ad_type: video
      - blobs: {creative_asset_url: "https://archive.org/download/folgers_10/folgers_10_512kb.mp4"}
        ad_id: sample-folgers-coffee
        brand_slug: folgers
        brand_name: "Folgers"
        source: archive.org
        source_url: "https://archive.org/details/folgers_10"
        headline: "Folgers Coffee"
        primary_text: "A vintage television commercial for Folgers Coffee."
        ad_type: video
      - blobs: {creative_asset_url: "https://archive.org/download/sanka_coffee/sanka_coffee_512kb.mp4"}
        ad_id: sample-sanka-coffee
        brand_slug: sanka
        brand_name: "Sanka"
        source: archive.org
        source_url: "https://archive.org/details/sanka_coffee"
        headline: "Sanka Coffee"
        primary_text: "A vintage television commercial for Sanka, a decaffeinated coffee."
        ad_type: video
      - blobs: {creative_asset_url: "https://archive.org/download/tide/tide_512kb.mp4"}
        ad_id: sample-tide-detergent
        brand_slug: tide
        brand_name: "Tide"
        source: archive.org
        source_url: "https://archive.org/details/tide"
        headline: "Tide"
        primary_text: "A vintage television commercial for Tide laundry detergent."
        ad_type: video
      - blobs: {creative_asset_url: "https://archive.org/download/ivory_soap/ivory_soap_512kb.mp4"}
        ad_id: sample-ivory-soap
        brand_slug: ivory
        brand_name: "Ivory"
        source: archive.org
        source_url: "https://archive.org/details/ivory_soap"
        headline: "Ivory Soap"
        primary_text: "A vintage television commercial for Ivory bar soap."
        ad_type: video
      - blobs: {creative_asset_url: "https://archive.org/download/kelloggs_variety_pak/kelloggs_variety_pak_512kb.mp4"}
        ad_id: sample-kelloggs-variety-pak
        brand_slug: kelloggs
        brand_name: "Kellogg's"
        source: archive.org
        source_url: "https://archive.org/details/kelloggs_variety_pak"
        headline: "Kellogg's Variety Pak"
        primary_text: "A vintage television commercial for the Kellogg's Variety Pak of breakfast cereals."
        ad_type: video
      - blobs: {creative_asset_url: "https://archive.org/download/post_grape-nuts_flakes/post_grape-nuts_flakes_512kb.mp4"}
        ad_id: sample-post-grape-nuts-flakes
        brand_slug: post
        brand_name: "Post"
        source: archive.org
        source_url: "https://archive.org/details/post_grape-nuts_flakes"
        headline: "Post Grape-Nuts Flakes"
        primary_text: "A vintage television commercial for Post Grape-Nuts Flakes breakfast cereal."
        ad_type: video
      - blobs: {creative_asset_url: "https://archive.org/download/amc_jeep/amc_jeep_512kb.mp4"}
        ad_id: sample-amc-jeep
        brand_slug: jeep
        brand_name: "Jeep"
        source: archive.org
        source_url: "https://archive.org/details/amc_jeep"
        headline: "AMC Jeep"
        primary_text: "A vintage television commercial for the AMC Jeep."
        ad_type: video
acceptance:
  max_cost_usd: 1
  steps:
    - name: ad-scenes-indexed
      kind: documents
      collection: ads-scenes
      expect: {min_documents: 8}
    - name: ad-copy-indexed
      kind: documents
      collection: ads-text
      expect: {min_documents: 8}
    - name: brand-scenes-answers
      kind: query
      retriever: brand-scenes
      inputs: {query: a cup of coffee}
      expect: {min_results: 1}
    - name: copy-search-answers
      kind: query
      retriever: copy-search
      inputs: {query: breakfast cereal}
      expect: {min_results: 1}

Read from the manifest, so a part it ships commented out never shows as ready.

ads-scenesCollectionDeploys ready

ads-textCollectionDeploys ready

brand-scenesRetrieverDeploys ready

copy-searchRetrieverDeploys ready

competitor-product-monitorRetrieverDeploys ready

creative-patternsClusterDeploys ready

ad-library-pullData sourceNeeds your input

ads-facesCollectionNot available yet

cast-identificationRetrieverNot available yet

What each part needs

ads-scenes: Deploys and processes documents as they arrive.
ads-text: Deploys and processes documents as they arrive.
brand-scenes: Deploys and answers queries once its collections hold documents.
copy-search: Deploys and answers queries once its collections hold documents.
competitor-product-monitor: Deploys and answers queries once its collections hold documents.
creative-patterns: Deploys and processes documents as they arrive.
ad-library-pull: You upload these files yourself. There is no connection to sync.
ads-faces: Face recognition is turned off on Mixpeek Cloud, so this face step does not deploy. It runs on a dedicated or customer-hosted deployment, where face_identity_extractor is served.
cast-identification: Face recognition is turned off on Mixpeek Cloud, so this face step does not deploy. It runs on a dedicated or customer-hosted deployment, where face_identity_extractor is served.

DocsMultimodal extractor Face identity extractor

Component	Role	Source	Today
multimodal_extractor@v1	Cuts each creative into scenes and embeds themRun with split_method scene, scene_detection_threshold 0.5 and time_split_interval 10, which is the configuration the measured numbers above came from.	Built in	Deploys ready
face_identity_extractor@v1	Detects, aligns and embeds every faceNot deployed on Mixpeek Cloud, where face recognition is off. On a dedicated or customer-hosted deployment: scrfd_2.5g detection, min_face_size 20, quality scoring on, embeddings normalised.	Built in	Not available yet
text_extractor@v1	Embeds the ad copyEmbeds each ad's primary_text as one chunk, which is what the manifest asks for: it sets no split_by, and the extractor default is none. To chunk, set split_by and a chunk_size counted in that splitter's own unit, so sentences with chunk_size 5 gives five-sentence chunks.	Built in	Deploys ready
brand-scenes	The retriever to run first: plain language in, ranked scenes out	Built in	Deploys ready

DocsCustomer-hosted overview

Mixpeek cloud

Apply the manifest and the namespace comes up with three collections and three retrievers.

Your own cloud

Same manifest against a customer-hosted deployment.

Permissions: A key with namespace create and collection create. The storage connection is a separate grant and the manifest ships it commented out, because apply live-tests a connection and would roll the whole namespace back on placeholder credentials.
Footprint: Scene extraction is the cost driver and is billed per minute of video. On the measured corpus, 271 source videos with a median length of 29 seconds produced 2,306 scenes and 8,168 face documents.

DocsRetrieval cookbook

Reference fixtures

None ship with this template. The measured corpus above is a live namespace rather than a fixture set, and there is no golden query list to score against yet.

Category	Rule
Scene detection	scene_detection_threshold 0.5 with a 10 second time split as the fallback. Lower it and single-take creatives stop producing one scene; raise it and cuts get merged.
Face quality	On a dedicated deployment, where the face step runs: min_face_size 20 pixels and detection_threshold 0.5, with quality scoring on so a low-confidence face is filterable rather than silently trusted.
Duplicate creatives	An archive that carries the same creative many times will produce clusters dominated by one ad. Collapse duplicates before presenting clusters as distinct themes.

Throughput

One brand-scenes execution over the measured corpus returned 24 documents in 1.09 seconds. That is a single observation rather than a distribution, and executions on the same namespace have also returned 502 and 503 under shard load, so treat about a second as the good case rather than a guarantee.

Does this search what people say in the ads?+

No. This template indexes what is on screen, who is in frame, and the ad copy you supply as a field. Spoken audio is not part of it, and the measured corpus carries no transcripts at all, so nothing on this page depends on speech.

What do the relevance scores look like?+

Cross-modal text-to-video similarity sits in a narrow band, roughly 0.11 to 0.13 on the measured corpus. That is normal for this kind of retrieval and it is why the template does not render a score as a percentage: an 11% label reads as a broken product when the ranking is working correctly. Use rank order, or a bar relative to the top result.

Why are faces a separate collection instead of a field on the scene?+

One scene can contain several faces and one face appears in many scenes, so attaching faces to scenes would force a choice about which face the scene is. Keeping them separate lets a face query return every appearance across the archive and a scene query stay about composition.

Do the clusters work out of the box?+

Clustering is a per-brand run rather than a step in the pipeline, and on the measured namespace 9 of 16 runs completed, 3 failed outright, 1 was in progress and 1 completed with errors. Plan for a brand's cluster run needing a retry, and read the cluster status before presenting themes as complete.

How much of the archive does copy-search reach?+

Ad copy is a field you supply on the object, not something the pipeline extracts from the video, so coverage is whatever your export carries. On the measured corpus it is 38 creatives out of 271, about 14%, with a median chunk of 184 characters. The retriever is correct and the index is the real multilingual one, and it is worth knowing that copy-search answers over the creatives that have copy rather than over the whole archive.

Can I point it at a public ad library instead of my own exports?+

The pipeline does not care where the video came from, only that the object carries a video URL and the copy. What changes is the shape of the media: the measured corpus is vertical ad creative, 5 to 80 seconds, which cuts into scenes readily. Longer single-take footage will produce one scene per file and the numbers on this page will not reproduce.