Find any scene in your video library.

    Mixpeek understands your video, image, audio, and documents, and returns the most accurate, timestamped results from the storage you already have. Every search teaches it: relevance keeps improving the more you and your agents use it.

    Create your account Talk to an engineer
    How it compounds
    Featured case study

    A performance-video agency maps the creative DNA of 170K+ scenes across 8 brands : search any moment, cluster what works, reuse the winners. See how

    Works with your DAMiconik·Mux+ more integrations →
    Ready to build?

    Explore the live demos above with zero setup: then start here when you want Mixpeek running on your own data.

    Bring your own vectors

    You already have embeddings

    Mixpeek Vector Store (MVS): an agent-native vector store that runs on your object storage. Dense, sparse, and BM25 search. From $25/mo.

    Start with your vectors Read MVS docs

    Connect your files

    You have raw files

    Managed indexing: point us at video, audio, or documents and we extract scenes, faces, OCR, transcripts, and embeddings. No pipeline to build. From $25/mo.

    VideoImagesAudioPDFsText
    Start with your files Talk to us
    Built for production retrieval
    $25per month minimum, usage-based
    $0.023per GB-month of vector storage
    $2per million queries
    Billionsvectors, scale on object storage
    Perception layer

    What the agent sees.

    Every object you index becomes structured, searchable features: faces and objects in a frame, layout regions in a document, speakers in audio. Those are the same features an agent queries, and joins across modalities.

    Hover or tap a card to preview the search it powers.

    Person walking past a storefront, with face, body, and handbag detectedPerson · 0.97Face · 0.95Handbag · 0.92Video · 00:04:12
    scene caption

    “A woman carrying a tan tote bag walks past a red storefront.”

    transcript · “…meet me at the corner in five.”

    search
    A printed resume with header, body, chart, and signature regions detectedHeaderChartBodySignaturePDF · resume.pdf
    layout extraction

    Header, body, charts, and signature detected as typed regions.

    OCR · 1 header · 3 sections · 1 signature

    search
    A studio microphone, with audio diarized into speaker segments
    Speaker A
    Speaker B
    Speaker A
    match · “limited time offer”
    speaker diarization

    Who spoke when, aligned to the transcript and the timeline.

    audio · 2 speakers · matched at 00:01:30

    search
    View the extractors that produce these features
    Cross-modal joins

    One query across every modality.

    Real questions rarely fit one feature. “Find the moment our CEO said guidance while the slide read Q4 outlook” needs a face, a spoken phrase, and on-screen text to line up at the same instant.

    Mixpeek ties those features to the same object and timestamp, so an agent gets back the exact clip instead of three unrelated matches.

    See how multi-stage retrieval joins features
    One query, three features
    Video
    face_id = person_42
    Audio
    transcript contains “guidance”
    On-screen text
    slide reads “Q4 outlook”
    ⨝ join on object + timestamp
    earnings_call.mp4 · 00:12:040.93

    The CEO says “guidance” as the slide behind her reads “Q4 outlook.”

    matched onface_idtranscripton-screen text
    Discovered structure

    "What concepts exist in my data that nobody has labeled yet?"

    No keyword, example, or prompt can answer that: they all assume you already know what you're looking for. Mixpeek clusters what belongs together on its own, so a natural hierarchy surfaces instead of a flat pile of tags: a taxonomy built from your data, organized around what your business actually cares about.

    That taxonomy is your ground truth, and it feeds back. Every search, every correction sharpens the features, the clusters, and the relationships. Your competitors' metadata decays. Yours compounds.

    How the compounding loop worksExplore clustering
    Features → clusters → taxonomy
    unboxing moments · 214night driving · 88product on white · 342no labels provided
    discovered taxonomy
    • Creative moments644
    • └Unboxing214
    • └Hands-on close-up121
    • └Reveal + reaction93
    • └Night driving88
    • └Product on white342

    every search + correction feeds back → sharper features, tighter clusters, truer taxonomy

    Experience memory

    "Have I solved this before?"

    A computer-use agent hits a screen it cannot get past. Somebody unsticks it once. Thirty days later the same wall appears on a redesigned page, under a different account, and the agent recovers on its own by retrieving what worked last time. Not a summary of the conversation: the screen, the action, the tool result, and the transition that followed.

    The step it found

    Retrieval returns the packet itself, so the agent can read the evidence rather than trust a paraphrase of it.

    {
  "observation_id": "obs_run_42_step_08",
  "run_id": "run_42",
  "step": 8,
  "task": "reset account password",
  "action_before": { "type": "click", "target": "Save" },
  "visual_state": {
    "caption": "Settings page, save button disabled",
    "ocr": ["Current password", "New password", "Save"],
    "ui_elements": [
      { "role": "button", "text": "Save", "enabled": false }
    ]
  },
  "tool_outputs": [
    { "tool": "validate_password_policy",
      "result": "missing_special_character" }
  ]
}
    Closest successful precedent
    run_42 / step_0830 days earlier
    Why it matched
    • Screen · a settings page with the save control greyed out
    • OCR · "Current password", "New password", "Save"
    • UI state · button "Save", enabled: false
    • Tool output · validate_password_policy → missing_special_character
    What happened next
    1. step_09added a special character to the new password
    2. step_10Save became enabled
    3. step_11account updated

    Ask the agent why it did that and the answer is a citation, not a recollection: the run, the step, the screen it matched, and the transition it copied.

    How to build thisTesting whether the recall holds upGiving the agent one tool instead of a vector database
    Templates

    Start from a complete namespace.

    Use a template when the outcome is clear and the stack should arrive together: storage sources, extraction stages, retrievers, evaluation rules, and deployment notes.

    Namespace

    Sources, buckets, collections, and retrievers ship together.

    Manifest

    One YAML file mirrors the rendered flow diagram.

    Evaluation

    Thresholds and review rules are attached to the template.

    Browse templates Retriever templates
    Two people filming a piece to camera with a ring light on a table
    Advertising1 / 8

    Creative DNA for Ad Archives

    An ad archive you can query by what is on screen. Every creative is cut into scenes at ingest, every face is indexed, 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.

    2,306

    scenes across 271 source videos

    8,168

    face documents from the same archive

    Open template
    Pexels / ANTONI SHKRABA production
    Quickstart

    One install. Two paths.

    Most retrieval stacks mean gluing together a vector DB, a file pipeline, and an agent layer. Mixpeek is one install with two ways in.

    Install
    pip install mixpeek
    MVS quickstart

    Bring embeddings

    Create MVS namespace Read MVS docs
    # pip install mixpeek
from mixpeek import Mixpeek

mp = Mixpeek(api_key="YOUR_KEY")

mp.namespaces.documents.upsert(
    namespace="agent_memory",
    documents=[{
        "id": "clip_001",
        "dense_embedding": embedding,
        "metadata": {"source": "s3://media/clip_001.mp4"}
    }]
)

results = mp.namespaces.documents.search(
    namespace="agent_memory",
    queries=[{"vector": query_embedding, "top_k": 10}]
)
    01BYO embeddings
    02Object storage
    03Agent retrieval
    File types × search-by features

    Pick your file types. Choose what to search by.

    Video, image, audio, documents, text, or web: connect a bucket, pick the features you want to search by, and these pipelines run as they are. Every one is documented and open source in the extractor cookbook.

    View all extractors
    Founded by engineers from
    MongoDBBerkeleyNVIDIAEtsyAmazon Web ServicesEquinixIAB Tech Lab
    Enterprise-ready
    SOC 2-ready Encrypted at rest BYO-Cloud Self-hosted option Audit trails SSO / SAML
    Pricing

    From $25/mo. Usage-based everything.

    Two products, one model: a monthly minimum that acts as a floor, with usage above it billed at the same transparent rates. MVS is priced by the vector, Managed by the object.

    MVS
    $25/mo minimum

    Bring your own embeddings and pay by the vector. Dense, sparse, and BM25 search on your own object storage. Build starts at $25/mo with up to 1M vectors; Scale ($250/mo) covers 25M.

    Start with MVS
    Managed
    $25/mo minimum

    Bring raw objects and pay by the object: credits at $0.001 cover extraction, embedding, indexing, enrichment, and retrieval. Build covers 100K objects/mo; Scale ($250/mo) covers 1M.

    Start with Managed
    Enterprise
    Custom

    Dedicated infrastructure, self-hosted options, SSO, SLA, security reviews, and hands-on architecture support.

    Talk to us
    See full pricing
    FAQ

    Common questions.

    Do I have to move my data?

    No. Mixpeek reads from your existing S3, GCS, R2, Azure, or S3-compatible bucket. Your storage stays the system of record, and nothing leaves your cloud.

    How fast is retrieval?

    Hybrid queries (dense, sparse, and BM25) return in well under 100ms p95, even with vectors persisted on object storage rather than held in RAM.

    Do I need embeddings to start?

    No. Bring your own vectors with MVS, or point Managed at raw files and it generates embeddings and features for you.

    What can Managed extract?

    Faces, scenes, transcripts, OCR, labels, and embeddings from video, images, audio, PDFs, and documents, all indexed at the object level.

    Can I self-host?

    Yes. Deploy in your own cloud (BYO-Cloud) with encryption, role-based access, SSO, audit trails, and namespaces. We hold no SOC 2 or HIPAA certification today; see mixpeek.com/trust.

    How does pricing work?

    Both MVS and Managed start at $25/mo minimum. Usage counts toward the minimum: pay the greater of metered usage or the floor. MVS bills storage + queries; Managed bills in credits covering extraction, embedding, indexing, and retriever execution.