
Mixpeek
Mixpeek is a multimodal warehouse for AI that searches, classifies, filters, ranks and moderates video, images, audio and documents where they already live in your storage, exposing the extracted data through one API.
About Mixpeek
Mixpeek
Mixpeek reads video, images, audio and documents in existing S3, GCS, R2, Azure or S3-compatible buckets and extracts faces, scenes, objects, transcripts, OCR text, labels, embeddings and document layout. Applications can then search, classify, filter, rank and moderate that data through one API. The site describes Managed indexing (connect a bucket and extraction runs automatically), Managed templates that deploy a pipeline in one click, Bring-your-own-models with 400+ models, and Mixpeek Vector Store (MVS) for bringing your own vectors with dense, sparse and BM25 search on object storage. The page states hybrid queries return in well under 100ms p95, that storage remains the system of record, and that BYO-Cloud and self-hosted deployment are available. The page states it holds no SOC 2 or HIPAA certification today. Integrations listed include S3, GCS, R2, Mux, Backblaze B2, Iconik and LangChain, among others.
Key features
- Managed indexing that extracts scenes, faces, text and transcripts from connected buckets
- Feature extractors for faces, scenes, transcripts, OCR and fingerprints
- Mixpeek Vector Store (MVS) with dense, sparse and BM25 search on object storage
- Retrievers that compose multi-stage search with filter, join and rerank
- Access to 400+ models, or running your own models on Enterprise
- Cross-modal joins that align what is seen, said and shown at the same moment in a video
- Automatic clustering and taxonomy discovery for unlabeled data
- Prebuilt templates that deploy a working pipeline in one click
- Integrations with S3, GCS, R2, Mux, LangChain and MCP
- Python SDK installable with pip install mixpeek
- Enterprise options including BYO-Cloud, self-hosting, SSO/SAML, audit trails and encryption at rest
Best suited for
- Ad creative analysis and finding winning video hooks
- Video moderation and brand mark matching
- Searching video, image, audio and document archives
- Document intelligence on PDFs and scans, including layout and tables
- Fleet safety, sports performance, manufacturing QA and similar video search use cases shown on the page
Pricing
From $25/mo. Usage-based. MVS starts at $25/mo plus usage (Build covers up to 1M vectors; Scale is $250/mo). Managed starts at $25/mo plus usage (Build includes $10 processing and 100K objects; Scale is $250/mo). Enterprise is custom pricing.
Discussion
Used Mixpeek? Share what worked and what didn’t.
No comments yet. Start the conversation.