Mixpeek
Mixpeek is an intelligent file store that utilizes advanced extraction, indexing, and searching technology. With a single API, developers c...
Last verified:
What is Mixpeek?
Mixpeek is a multimodal data warehouse and retrieval infrastructure for AI agents that transforms unstructured video, audio, images, PDFs, and documents into instantly searchable, structured data. It provides a single retrieval API over all file types, operating on the object storage you already use (S3, GCS, B2, Azure, R2, MinIO) without requiring data migration. The platform automatically extracts features like faces, objects, scenes, OCR text, transcripts, and embeddings from any file type through custom AI extractors and retrieval pipelines.
Key features include hybrid search (dense, sparse, and BM25), multi-stage agent retrieval pipelines, face search with ArcFace embeddings, scene detection with CLIP/Gemini, Whisper transcripts with embeddings, visual embeddings (SigLIP/Vertex AI), OCR for scanned PDFs, and structured extraction. Mixpeek offers both MVS (Mixpeek Vector Store) for bring-your-own-embeddings use cases and Managed for raw file ingestion with automatic extraction. The platform supports LangChain retrievers, MCP servers, REST API, and Python SDK integration.
Mixpeek is designed for enterprises and developer teams in advertising, media & entertainment, e-commerce, education, and AI product development. It's ideal for talent search across video ads, visual search across artwork collections, movie taste personalization, face search across video corpora, ad performance analysis, and building privacy-safe scalable applications. The platform is SOC 2 Type II certified, HIPAA-ready, offers self-hosted deployment options, and provides production controls like usage limits, audit trails, and namespaces.
Mixpeek pricing
Pricing model: Free
Mixpeek offers three pricing tiers: Free ($0/month) includes 1K credits/month, 3 collections, 1 namespace, basic extractors, free searches and retrievals, and 10 requests/minute rate limit. Pro ($99/month) includes 25K credits/month, 50 collections, 10 namespaces, all extractors, webhooks, batch processing, role-based access control, email support, and 60 requests/minute rate limit, with overage at $0.001/credit. Enterprise (custom pricing) starts at $2,500/month platform fee for dedicated single-tenant infrastructure, plus 20% management fee on cloud compute costs, including unlimited credits/collections, dedicated Slack support, forward-deployed engineering, custom integrations/plugins/extractors, single-tenant deployment, performance tuning, custom model training/fine-tuning, dedicated support team with SLA, SSO, audit logs, and compliance. Credits cover extraction, embedding, indexing, enrichment, and retriever execution. Multimodal Extractor costs $0.05/minute video, $0.005/image, $0.002/1K text tokens. Face Identity Extractor costs $0.005/image processed or $0.005/per face detected. Web Scraper costs $0.005/page crawled, $0.001/code block embedded, $0.002/image embedded. 1 credit = $0.001. MVS starts with 1M vectors free forever.
Mixpeek pros
- 1M vectors free forever with no expiration on MVS tier
- 10x lower cost compared to in-memory vector databases
- Hybrid retrieval under 100ms p95 even with object storage persistence
- Billions of vectors scale on object storage without RAM requirements
- No data migration needed - reads from existing S3, GCS, Azure buckets
- Single retrieval API covers video, images, audio, PDFs, and text
- Automatic feature extraction: faces, scenes, OCR, transcripts, embeddings
- Multi-stage retrieval pipelines with search, filter, join, rerank capabilities
- Deterministic, auditable traces for every retrieval operation
- SOC 2 Type II certified and HIPAA-ready for enterprise compliance
- Self-hosted and BYO-Cloud deployment options for data sovereignty
- Integrations with Mux, Backblaze B2, Iconik, LangChain, and MCP
- Webhooks and batch processing available on Pro tier
- Role-based access control and namespace isolation for tenants
- Custom model training and fine-tuning available on Enterprise tier
- Free searches and retrievals included in all pricing tiers
- Inference caching for repeated calls to same feature URI
- Supports 60 seconds to first query with MVS quickstart
Mixpeek cons
- Free tier limited to 1K credits per month and 3 collections
- Free tier rate limit of only 10 requests per minute
- Maximum 50MB file size for search operations
- Maximum 2GB file size for indexing operations
- Pro tier at $99/month may be expensive for small teams
- Credit-based pricing can become costly at high volume without discounts
- Enterprise requires $2,500/month minimum platform fee plus 20% compute management fee
- Some premium extractors like Gemini Embedding 2 cost $0.05 per video minute
- Non-native file formats auto-convert to JPEG or MP4, losing original format
- Free tier includes only single namespace and basic extractors only
Frequently asked questions about Mixpeek
Do I have to move my data to use Mixpeek?
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. You connect any object store, point Mixpeek at it, and every file becomes searchable by what's inside it without migration or code changes.
How fast is retrieval with Mixpeek?
Hybrid queries combining dense, sparse, and BM25 search return in well under 100ms p95, even with vectors persisted on object storage rather than held in RAM. The platform delivers 100ms p95 hybrid retrieval at scale.
Do I need to provide embeddings to start using Mixpeek?
No. You have two options: Bring your own vectors with MVS (Mixpeek Vector Store), or point Managed at raw files and Mixpeek generates embeddings and features for you automatically. MVS lets you skip extraction entirely and search instantly with 1M vectors free and 60 seconds to first query.
What can Managed extract from my files?
Managed extracts faces, scenes, transcripts, OCR, labels, and embeddings from video, images, audio, PDFs, and documents, all indexed at the object level. Specific features include face embeddings (ArcFace 512D), scene descriptions (Gemini), visual embeddings (Vertex AI 1408D), transcripts (Whisper), transcript embeddings (E5-Large 1024D), keyframes, visual embeddings (SigLIP 768D), OCR text, descriptions, structured extraction, text chunks, and text embeddings.
Can I self-host Mixpeek in my own cloud?
Yes. You can deploy in your own cloud with BYO-Cloud (Bring Your Own Cloud) featuring SOC 2 and HIPAA-ready controls, SSO, audit trails, and namespaces. Enterprise tier includes single-tenant deployment options or deployment in your cloud with dedicated infrastructure.
How does the credit system work for pricing?
Each feature extractor charges credits based on what it processes: minutes of video, number of images, text tokens, document pages, detected faces, and other billing units. 1 credit equals $0.001. Credits are deducted as extractors process data with pay-per-unit processed pricing. Managed searches and retrievals are included and always free regardless of credit usage.
What's the difference between MVS and Managed?
MVS (Standalone) is a pure vector store where you bring your own embeddings and pay for object storage, hot cache, queries, and writes. Managed includes Mixpeek's extraction pipeline where you send raw files (video, images, PDFs) and Mixpeek handles chunking, embedding, and indexing using credits. You can use MVS as just a vector store without extraction and add extraction pipelines later without rebuilding your retrieval layer.
Are there long-term commitments for Mixpeek plans?
No. All self-serve plans (Free and Pro) are billed monthly with no long-term commitments. You can upgrade, downgrade, or cancel at any time. Enterprise pricing is custom based on committed usage, deployment model, and support requirements.
What file types does Mixpeek support?
Images: JPEG, PNG, HEIC (auto-converts to JPEG), WebP (auto-converts to JPEG), TIFF (auto-converts to JPEG), BMP (auto-converts to JPEG), GIF (processed as image or video). Videos: MP4 (native support), MPEG (auto-converts to MP4), QuickTime MOV (auto-converts to MP4), AVI (auto-converts to MP4), WMV (auto-converts to MP4). Also supports PDFs, audio files, and text documents. Format conversion happens automatically during processing.
Are there volume discounts for high usage?
Yes. On Pro tier, usage beyond included 25K credits is billed at $0.001/credit with volume discounts at higher usage levels. Enterprise (single-tenant) customers get custom pricing based on committed usage, deployment model, and support requirements. Contact sales for a custom quote on enterprise pricing.