Search every second of your video archive
Production-grade video search infrastructure for companies with private archives. Deploy in your cloud or ours. SSO, SLA, custom indexing, dedicated support.
0%
Uptime SLA
0ms
Avg. search latency
0+
Hours indexed daily
0/7
Dedicated support
Your video archive is dark data
Companies are accumulating thousands of hours of video — training, all-hands, sales calls, customer recordings, lectures, productions — and almost none of it is actually queryable.
Knowledge is locked in pixels
Every meeting decision, training explanation, and customer insight is trapped behind a 60-minute scrub bar. Your team's institutional memory is technically present but practically unreachable.
AI agents can't see video
Your Claude Code, Cursor, and internal agents have web search, code search, and document search — but no way to ground answers in video. Your video archive is invisible to your AI stack.
Building it yourself is a year
Pipelines for transcription, frame extraction, multimodal embedding, hybrid retrieval, reranking, indexing, and a maintainable API surface — typical buildout is 6–12 months with two senior engineers, plus ongoing maintenance.
Who uses Cerul Enterprise
Any organization with a serious video corpus and a real need to query it.
Education & Training
Online learning platforms, corporate training, certification
- “Find every moment a lecturer explains backpropagation”
- “Surface the exact second a compliance topic is covered”
- “Answer learner questions with video citations”
Media & Publishing
Broadcasters, news organizations, video archives
- “Locate B-roll matching a news story in seconds”
- “Search decades of footage by spoken topic or on-screen visual”
- “Surface clips for content licensing requests”
Video SaaS
Meeting recorders, sales call platforms, video CMS
- “Add semantic search across all your customers' recordings”
- “Power AI summaries with grounded video citations”
- “Make meeting libraries instantly queryable”
Stock & Footage
Stock libraries, archive houses, production companies
- “Index millions of clips by visual + spoken content”
- “Replace tag-based search with semantic retrieval”
- “Enable natural-language clip discovery for creators”
Internal Knowledge
All-hands archives, Zoom recordings, training videos
- “Search every all-hands meeting your company has ever had”
- “Surface decisions made in past planning sessions”
- “Onboard new hires with searchable historical context”
Legal & Compliance
Depositions, recorded calls, regulatory archives
- “Search depositions for testimony on a specific topic”
- “Find compliance discussions across recorded calls”
- “Build defensible records of what was said and when”
What Enterprise gives you
Production capabilities beyond what we ship in our hosted public API.
Visual + speech + OCR fusion
Index what is said, what is shown on screen, and what is written in slides or code — all aligned to a single timeline.
Multimodal embeddings
State-of-the-art encoders (SigLIP, Voyage, or your choice) power retrieval that understands meaning, not just keywords.
Tuned reranker
Our domain-aware reranking model boosts relevance by 30-50% over standard hybrid retrieval — only available in Enterprise and Cloud.
Agent-ready API
JSON-native REST endpoints with stable contracts. Designed for Claude Code, Cursor, autonomous agents, and your internal tools.
Private indexing
Your videos and embeddings never leave your environment. Index runs in your VPC, on-prem, or in our dedicated tenant.
Custom embedding models
Bring your own fine-tuned encoder for domain-specific accuracy on medical, legal, scientific, or technical content.
Bulk ingestion
Process millions of hours of historical video with parallel pipelines. We've benchmarked at 10,000+ hours indexed per day per cluster.
Incremental sync
New content gets indexed within minutes of upload. Sources include S3, Azure Blob, GCS, Zoom Cloud, YouTube channels, RSS, and custom integrations.
Three ways to deploy
Run Cerul where your data needs to live. We meet you where you are.
Dedicated SaaS
Fastest to deploy
We run everything in a dedicated tenant on our infrastructure. Single-tenant isolation, dedicated resources, your data stays in your region.
Best for
Teams that want managed infrastructure with isolation guarantees.
- Single-tenant deployment
- Region selection (US, EU, APAC)
- Managed updates and scaling
- 99.9% uptime SLA
- Daily-to-go-live in 2 weeks
Private Cloud
Your AWS / GCP / Azure
We deploy Cerul inside your cloud account. Your network, your IAM, your audit logs. We provide updates, support, and tuning.
Best for
Regulated industries, sensitive workloads, existing cloud footprint.
- Runs in your cloud account
- Your VPC, IAM, and KMS
- Terraform / Pulumi modules included
- We handle updates remotely
- Daily-to-go-live in 4–6 weeks
On-Premise
Air-gapped capable
Full on-prem deployment for environments without cloud access. Includes Helm charts, offline model bundles, and white-glove install.
Best for
Government, defense, sovereign data, fully air-gapped networks.
- Kubernetes / bare metal
- Offline model bundles
- Air-gapped install path
- On-site or remote install support
- Daily-to-go-live in 8–12 weeks
Built for enterprise security review
Your data never leaves your control. Standard controls, standard integrations.
Our foundational promise
Your data never trains any model.
Customer videos, transcripts, embeddings, and queries are never used to train any Cerul model or third-party model. We do not share customer data with model providers for training. This is binding — written into our DPA.
SOC 2 Type II
Audit in progress; controls inherited from our cloud infrastructure today.
SSO / SAML / SCIM
Integrate with Okta, Google Workspace, Azure AD, JumpCloud, OneLogin.
Audit logs
Every query, ingestion event, and admin action logged and exportable.
Encryption
TLS 1.3 in transit, AES-256 at rest. BYOK with AWS KMS / Azure Key Vault.
Role-based access
Fine-grained RBAC for teams, sources, and indexes.
Data residency
Choose region; full data isolation between tenants in Dedicated SaaS.
The TCO is not close
We've been building video search in production for years. Most companies underestimate the cost of doing it themselves by an order of magnitude.
| Dimension | Build it yourself | Cerul Enterprise |
|---|---|---|
| Engineering time | 6+ months, 2 senior engineers | 2–8 weeks (depending on deployment) |
| Upfront cost | $300k – $600k (loaded) | Setup fee + annual contract |
| Ongoing maintenance | 0.5–1 FTE forever | Included in support contract |
| Model upgrades | Each upgrade = re-architecture project | Handled by us; you stay current |
| Quality (retrieval relevance) | Baseline open-source models, no tuning | Tuned reranker + multimodal fusion |
| Time to first result | 9–12 months | Days to weeks |
| Risk | Model selection, scale, ops all on you | Production-grade, with SLA |
How we work with you
Discovery call
30 minutes. We learn your video corpus, search use case, security requirements, and deployment constraints.
Scoped pilot
We index a representative sample (typically 100–1,000 hours) and you evaluate retrieval quality against your real queries.
Production deployment
Choose deployment model (SaaS, Private Cloud, or On-Prem). We deploy, integrate auth and storage, and run your full corpus through the pipeline.
Ongoing partnership
Dedicated support, regular model upgrades, custom feature work, and quarterly business reviews. You scale, we maintain.
Try Cerul self-hosted first
Cerul App is our open-source, self-hostable video search daemon. Same pipeline, same retrieval architecture — without the tuned reranker, custom models, and enterprise support. A great way to evaluate fit before talking to sales.
Get started with the OSS daemon
# Self-host on your machine
docker run -d \
-p 7777:7777 \
-v ./data:/data \
cerulai/cerul-app
# Open http://localhost:7777Frequently asked questions
How is Enterprise different from Pro?
Pro is a self-serve subscription against our public hosted API and corpus. Enterprise is a private deployment of the full Cerul stack — your own indexes, your own data, your own infrastructure (in your cloud, ours, or on-prem). Enterprise also unlocks the tuned reranker, custom embedding models, SSO/SAML, audit logs, and a dedicated support channel.
Can we deploy Cerul in our own AWS / GCP / Azure account?
Yes. Private Cloud deployment runs Cerul entirely inside your cloud account using Terraform / Pulumi modules we provide. Your VPC, your IAM, your KMS. We handle updates remotely via your standard CI/CD pipelines.
Do you support air-gapped or on-premise deployments?
Yes. We provide Helm charts, offline model bundles, and white-glove install support. Common for government, defense, regulated banking, and sovereign data environments.
What kind of video sources can you ingest?
Out of the box: S3, Azure Blob, GCS, local filesystem, YouTube channels, RSS feeds, Zoom Cloud Recordings, Loom, Vimeo. Custom integrations available — we've built ingestion for sales call platforms (Gong, Chorus), meeting recorders (Otter, Fathom), and proprietary DAMs.
What about ongoing model upgrades?
Embedding and reranking models improve every 6–12 months. We re-embed your corpus in the background as part of your support contract — you get the quality upgrade without any work on your side. We never break query compatibility.
Can we fine-tune the embedding model on our data?
Yes. For domain-specific corpora (medical, legal, scientific, technical), we offer custom embedding model training as an add-on. Typical improvement is 20–40% retrieval accuracy on domain queries.
How does pricing work for Enterprise?
Pricing is based on three factors: total hours indexed, query volume, and deployment model. Most customers land between $30k–$300k/year. We start with a scoped pilot to validate fit before any long-term commitment.
What's the time to first production result?
Dedicated SaaS: 2 weeks. Private Cloud: 4–6 weeks. On-Premise: 8–12 weeks. Pilots typically start within a week of a signed mutual NDA.
Ready to make your video archive searchable?
30-minute discovery call. No slide deck. We learn your corpus, your use case, your constraints — and tell you straight whether Cerul is a fit.
Or write to panda@cerul.ai