AI vision for the cameras you already own.
Ocelli adds detection, rules, alerts, search and reports to your existing CCTV. It installs on one computer on your network, keeps every frame inside your building, and is priced per site, not per camera.
Alert me if a person is in the Loading Bay
between 20:00 and 06:00 for more than 30s,
unless they are wearing a hi-vis vest.
Person in restricted zone after hours for 34s. No hi-vis detected.
Recording is not monitoring
A 20-camera site records 480 camera-hours a day. Nobody watches it live, so footage gets reviewed only after something has already gone wrong.
Footage is not searchable
Finding one clip means scrubbing a timeline by hand. Investigations get dropped because they cost more staff time than the loss.
Motion alerts get switched off
Pixel-difference detection fires on rain, headlights and animals. Alert fatigue is the most common reason analytics deployments fail.
Adding AI usually means new cameras
Cloud analytics vendors bundle their own hardware. A 20-camera site pays for equipment it does not need, on top of annual per-camera licences.
Compliance has no product attached
CCTV footage containing faces is regulated personal data under DPDP and GDPR. Retention limits, deletion and access logs are handled today with a consultant and a spreadsheet.
Mid-size sites are not served
Analytics is sold as a bespoke project. A 12-camera cold store or a 30-camera school gets a quote it will not sign, or a free tool with no support.
Three components
Ocelli Edge
The detection runtime. Pulls camera sub-streams, runs the model, tracks objects and applies your rules. Sends nothing off site.
Ocelli Hub
Live view, events, zones, rules, users, dashboards, reports and the REST API. Runs on the same machine or your own server.
Ocelli Link
Cloud relay for multi-site fleets and remote access. Opt-in per feature. Leave it off and Ocelli runs fully offline.
Glance, Focus, Recall
Running a detector on every frame of every camera is what makes AI video expensive. Ocelli spends compute in three stages instead.
Glance
A low-cost gate watches every camera on a downscaled image and looks for meaningful change against a per-camera noise profile.
Focus
Only candidate frames reach the detector. Objects are tracked between frames, so rules can use dwell time, direction and sequence.
Recall
Events are indexed locally with snapshot and clip, so they can be searched, shared and reported on later.
What Ocelli does
Detection and tracking
People, vehicles and objects, confirmed across multiple frames before an event is raised. Every event stores confidence, boxes and the object track.
Zones and spatial rules
Polygons and lines drawn on the camera image: restricted areas, line crossing, dwell thresholds, per-zone schedules and privacy masks applied before inference.
Rules in plain language
Write the rule as a sentence. Ocelli compiles it into a deterministic rule you can inspect and version, and previews how often it would have fired.
Alerts on eight channels
WhatsApp, SMS, email, Telegram, Slack, mobile push, signed webhooks and MQTT. Per-rule severity, routing, throttling, quiet hours and escalation.
Events and incident workflow
Snapshot, clip, the rule that fired and the object track on every event. Acknowledge, assign, comment, resolve, export or share the clip.
Search your footage
A local index turns plain language into clips, such as "van at the gate last weekend". Runs offline. No cloud model unless you enable one.
Dashboards and reports
Per-industry KPIs, trends by camera, zone and hour, scheduled PDF reports, event export and system health reporting.
Tuning that compounds
Marking an alert as a false alarm feeds threshold tuning and zone-specific negatives. Your frames stay on your hardware.
Users, roles and multi-site
Role-based access down to individual cameras, SSO on higher editions, an audit trail of who viewed or exported footage, and a rollup across sites.
Privacy controls
Faces blurred before storage, short retention defaults with real deletion, every view logged, and camera masking for areas you may not record.
Camera and fleet health
FPS and dropped-frame monitoring, auto-reconnect with backoff, remote diagnostics, bulk camera import and config cloning between sites.
API and integrations
REST API for cameras, zones, rules and events. HMAC-signed webhooks and MQTT publish for building management and SCADA systems.
Configured for your type of site
Select your industry during setup. Ocelli loads the zones, detection classes, alert rules, severities and dashboard KPIs for that kind of site. Eighteen packs are included.
One codebase, six hardware profiles
The installer detects the machine, selects the runtime and model size, and reports how many cameras it can carry. Same interface and same rules on every tier.
| Tier | Reference hardware | Accelerator | Runtime | Cameras | Power |
|---|---|---|---|---|---|
| T0 Nano | Raspberry Pi 5 8 GB + Hailo-8L | 13 TOPS | HailoRT | 6–10 | 10 W |
| T1 Base Recommended | x86 mini PC (N100/N305), 16 GB | iGPU + NPU | OpenVINO | 10–16 | 10–15 W |
| T2 Value | RK3588 SBC (Orange Pi 5, ROCK 5) | 6 TOPS NPU | RKNN | 6–10 | 5–8 W |
| T3 Pro | NVIDIA Jetson Orin Nano Super | 67 TOPS | TensorRT | 30–50 | 15–25 W |
| T4 Server | x86 server + RTX / T4 / A2 | Discrete GPU | TensorRT | 60–150 | 70 W+ |
| T-CPU | Any x86_64, no accelerator | None | ONNX Runtime | 2–4 | Varies |
Camera counts are per machine at a useful detection rate, and are confirmed against your own streams during the pilot.
How many cameras do you have?
Indicative edition and hardware tier for that camera count.
Deployed in a day, tuned in two weeks
No cabling, no camera procurement, no integration project. Software installs in under an hour. The rest of day one is zones and rules.
Site review
We review your camera list, resolutions, recorder and network. Output: hardware tier, cameras per machine, and the first rules to write.
Install
One installer on Linux, Windows or Docker. It detects the hardware, selects the runtime and model, and generates its own admin credentials.
Add cameras
ONVIF discovery finds cameras on the network. Others are added by RTSP URL. Credentials are encrypted at rest. Detection uses the sub-stream, so your recorder is untouched.
Select industry
Choose one of 18 packs. Zone templates, detection classes, alert rules and the KPI dashboard are applied. You adjust zones on the live image.
Tune
Rules are previewed against recent footage before going live. False alarms are one tap. Thresholds, dwell times and schedules are tightened over two weeks.
Handover
Users and roles, alert routing and escalation, scheduled reports, retention policy, signage templates and training on the incident workflow.
What you provide
- Existing IP cameras on RTSP, RTMP, HTTP/MJPEG or ONVIF. Any brand.
- A sub-stream enabled per camera, typically 640×360 or D1.
- One machine on the same network. We specify it; you buy it locally.
- Camera credentials and network access from that machine.
- One person who owns alerts.
Detection, rules, the interface and local alerts work with no internet connection.
Single site
One machine runs Edge and Hub together. Staff open it in a browser on the local network or over your VPN.
Multi-site
An Edge machine per site handles detection locally and survives a WAN outage. A central Hub gives one event list, dashboard and report set across locations.
Through an integrator
Partners get a fleet console across clients, remote health monitoring, bulk provisioning, white-label branding and reseller margin.
How the options differ
| Approach | Use your cameras | Rules & roles | Reports & compliance | Works offline | Pricing model |
|---|---|---|---|---|---|
| Cloud AI camera vendors | ✗ | ✓ | ✓ | ✗ | Per camera, per year, plus hardware |
| Enterprise VMS with analytics | ✓ | ✓ | ✓ | ✓ | Licence plus integrator project |
| Open-source NVR | ✓ | ✗ | ✗ | ✓ | Free, self-supported |
| Project-based analytics | ✓ | Partial | Partial | ✓ | Bespoke project quote |
| Ocelli | ✓ | ✓ | ✓ | ✓ | Per site |
Categories reflect the standard commercial model of each type of vendor.
Per site, not per camera
Adding a camera does not trigger a purchase decision. You pay for the site, up to your edition's camera limit.
Free
- Detection engine
- Zones and basic rules
- Local web interface
- Email and Telegram alerts
- All 18 industry packs
- 7-day event history
- Community support
One-time setup fee applies. See below.
Pro
- Everything in Free
- Plain-language rule engine
- WhatsApp, SMS, push, Slack, webhooks
- Event clips and sharing
- Scheduled PDF reports
- Retention and blurring controls
- Escalation and on-call routing
- Email support
Business
- Everything in Pro
- Multi-site dashboard
- RBAC and SSO
- REST API and MQTT
- Footage search
- Per-site tuning loop
- Compliance pack
- 8×5 support
Enterprise
- Everything in Business
- Departments and focal persons
- Campaign and turnaround windows
- Signed audit reports
- On-premise high availability
- Custom model training
- Integrations and SLA
- Named account manager
Free for up to 4 cameras
The same detection engine as the paid editions, on up to four cameras at one site, with no time limit. Commercial use inside your own business is permitted.
Included
- The full detection engine
- Up to 4 cameras at one site, any brand
- Zones, basic alert rules and camera schedules
- Local web interface, live view and event list
- Email and Telegram alerts
- All 18 industry packs
- 7 days of event history
- Face blurring on by default, on-premise processing
Not included
- More than 4 cameras or more than one site
- Plain-language rule engine
- WhatsApp, SMS, push, Slack and signed webhooks
- Event clips, clip sharing and scheduled reports
- Footage search and the tuning loop
- Roles, SSO, API and multi-site rollup
- Compliance pack and audit reports
- Warranty and support commitment
- Resale, hosting for third parties, or removing attribution
One-time setup fee
The software licence is free. Deploying it on your site is engineering work and carries a small one-time setup fee. It is not a subscription and not a per-camera charge.
- Hardware sizing and a check that your streams will run
- Installation and configuration, remote or on site
- Camera discovery, credentials and sub-stream setup
- Zone drawing and rule tuning against your footage
- Alert routing and a handover walkthrough
The amount is indicative and agreed with our sales team. It depends on your site, camera count, location and whether the work is remote or on site, and it is often credited against your first year on a paid edition. If your own IT team or integrator installs it, the software is still free.
What Ocelli does not do
No face recognition. Ocelli detects people and situations, not identities.
No emotion detection or biometric categorisation by any protected attribute.
No productivity surveillance. No idle-time tracking or individual performance scoring.
No silent cloud upload. On-premise by default. Cloud features are opt-in and labelled.
No unblurred storage by default. Un-blurring is a separate permission and is logged.
No indefinite retention. Retention defaults to 30 days, and deletion is real deletion.
No unlogged viewing. Every view of footage is recorded and visible to the account owner.
No hidden engine. The detection layer is auditable, and notice signage templates are included.
Common questions
Do I have to replace my cameras?
No. If a camera produces an RTSP, RTMP, HTTP/MJPEG or ONVIF stream, Ocelli can use it. Any brand, any age, mixed vendors on one site. We do not sell cameras.
Does my video leave the building?
Only if you enable a cloud feature. Detection, rules, alerts, the interface and search run on your own hardware. Outbound channels such as WhatsApp, SMS and email need connectivity, but they send an alert, not your footage.
How do you avoid false alarms?
Detections are confirmed across multiple frames rather than triggered by pixel change. Rules can require dwell time, direction, schedule and compound conditions. New rules are previewed against recent footage before being enabled, and marking an alert as a false alarm tightens the system. The target is under five alerts per site per day after two weeks of tuning.
What hardware do I need?
For most sites, one x86 mini PC on the same network as the cameras. Larger sites use a Jetson or a GPU server. The installer reports how many cameras the machine can carry, and we confirm it against your streams during the pilot. You buy the hardware locally and it stays your asset.
How long does deployment take?
A single-site install is normally completed in a day, of which the software takes under an hour. Allow two weeks of tuning before judging alert quality. Multi-site rollouts run one site at a time, cloning the first site's configuration.
What is the setup fee on the free edition?
The licence is free for up to 4 cameras at one site, with no time limit, including commercial use in your own business. The setup fee is a one-time charge for sizing, installation and tuning. It is not a subscription and not per camera. The amount is agreed with our sales team based on your site, camera count and location, and is often credited against your first year if you upgrade. If you install it yourself, the software is still free.
Can I upgrade from Free later?
Yes. It is the same software. A licence file raises the camera limit and unlocks the paid features in place. Cameras, zones, rules, users and event history are unaffected.
What happens if my licence expires?
Ocelli continues running for a 14-day grace period, so a renewal delay does not leave your site unmonitored. Your data remains yours and stays exportable.
Does it do face recognition?
No, by design. Ocelli detects people and situations, not identities, and faces are blurred before storage. This is a product policy, not a limitation we intend to remove.
Can my integrator install and support it?
Yes. Partners get a fleet console across clients, remote health monitoring, bulk provisioning, white-label branding and reseller margin. Contact us about the partner programme.
How does this help with DPDP and GDPR?
On-premise processing, face blurring by default, short retention with real deletion, access logging, camera masking and notice signage templates. Higher editions add a compliance pack with audit-ready reporting. Ocelli is a tool, not legal advice: you remain the data controller for your footage.
Which detections are available?
The core engine covers people, vehicles and common objects, combined with zones, tracking and time logic to produce intrusion, after-hours presence, loitering, line crossing, dwell, crowding, abandoned objects and vehicle events. PPE compliance, fire and smoke, falls and number-plate recognition are purpose-trained add-on packs. Confirm the packs you need with sales when scoping.
See it on your own cameras
Send us your camera list and what you need detected. You get the hardware tier, the camera count per machine, the first rules to write, and a firm price.
AI vision for the cameras you already own.
Ocelli adds detection, rules, alerts, search and reports to your existing CCTV. It installs on one computer on your network, keeps every frame inside your building, and is priced per site, not per camera.
Alert me if a person is in the Loading Bay
between 20:00 and 06:00 for more than 30s,
unless they are wearing a hi-vis vest.
Person in restricted zone after hours for 34s. No hi-vis detected.
Recording is not monitoring
A 20-camera site records 480 camera-hours a day. Nobody watches it live, so footage gets reviewed only after something has already gone wrong.
Footage is not searchable
Finding one clip means scrubbing a timeline by hand. Investigations get dropped because they cost more staff time than the loss.
Motion alerts get switched off
Pixel-difference detection fires on rain, headlights and animals. Alert fatigue is the most common reason analytics deployments fail.
Adding AI usually means new cameras
Cloud analytics vendors bundle their own hardware. A 20-camera site pays for equipment it does not need, on top of annual per-camera licences.
Compliance has no product attached
CCTV footage containing faces is regulated personal data under DPDP and GDPR. Retention limits, deletion and access logs are handled today with a consultant and a spreadsheet.
Mid-size sites are not served
Analytics is sold as a bespoke project. A 12-camera cold store or a 30-camera school gets a quote it will not sign, or a free tool with no support.
Three components
Ocelli Edge
The detection runtime. Pulls camera sub-streams, runs the model, tracks objects and applies your rules. Sends nothing off site.
Ocelli Hub
Live view, events, zones, rules, users, dashboards, reports and the REST API. Runs on the same machine or your own server.
Ocelli Link
Cloud relay for multi-site fleets and remote access. Opt-in per feature. Leave it off and Ocelli runs fully offline.
Glance, Focus, Recall
Running a detector on every frame of every camera is what makes AI video expensive. Ocelli spends compute in three stages instead.
Glance
A low-cost gate watches every camera on a downscaled image and looks for meaningful change against a per-camera noise profile.
Focus
Only candidate frames reach the detector. Objects are tracked between frames, so rules can use dwell time, direction and sequence.
Recall
Events are indexed locally with snapshot and clip, so they can be searched, shared and reported on later.
What Ocelli does
Detection and tracking
People, vehicles and objects, confirmed across multiple frames before an event is raised. Every event stores confidence, boxes and the object track.
Zones and spatial rules
Polygons and lines drawn on the camera image: restricted areas, line crossing, dwell thresholds, per-zone schedules and privacy masks applied before inference.
Rules in plain language
Write the rule as a sentence. Ocelli compiles it into a deterministic rule you can inspect and version, and previews how often it would have fired.
Alerts on eight channels
WhatsApp, SMS, email, Telegram, Slack, mobile push, signed webhooks and MQTT. Per-rule severity, routing, throttling, quiet hours and escalation.
Events and incident workflow
Snapshot, clip, the rule that fired and the object track on every event. Acknowledge, assign, comment, resolve, export or share the clip.
Search your footage
A local index turns plain language into clips, such as "van at the gate last weekend". Runs offline. No cloud model unless you enable one.
Dashboards and reports
Per-industry KPIs, trends by camera, zone and hour, scheduled PDF reports, event export and system health reporting.
Tuning that compounds
Marking an alert as a false alarm feeds threshold tuning and zone-specific negatives. Your frames stay on your hardware.
Users, roles and multi-site
Role-based access down to individual cameras, SSO on higher editions, an audit trail of who viewed or exported footage, and a rollup across sites.
Privacy controls
Faces blurred before storage, short retention defaults with real deletion, every view logged, and camera masking for areas you may not record.
Camera and fleet health
FPS and dropped-frame monitoring, auto-reconnect with backoff, remote diagnostics, bulk camera import and config cloning between sites.
API and integrations
REST API for cameras, zones, rules and events. HMAC-signed webhooks and MQTT publish for building management and SCADA systems.
Configured for your type of site
Select your industry during setup. Ocelli loads the zones, detection classes, alert rules, severities and dashboard KPIs for that kind of site. Eighteen packs are included.
One codebase, six hardware profiles
The installer detects the machine, selects the runtime and model size, and reports how many cameras it can carry. Same interface and same rules on every tier.
| Tier | Reference hardware | Accelerator | Runtime | Cameras | Power |
|---|---|---|---|---|---|
| T0 Nano | Raspberry Pi 5 8 GB + Hailo-8L | 13 TOPS | HailoRT | 6–10 | 10 W |
| T1 Base Recommended | x86 mini PC (N100/N305), 16 GB | iGPU + NPU | OpenVINO | 10–16 | 10–15 W |
| T2 Value | RK3588 SBC (Orange Pi 5, ROCK 5) | 6 TOPS NPU | RKNN | 6–10 | 5–8 W |
| T3 Pro | NVIDIA Jetson Orin Nano Super | 67 TOPS | TensorRT | 30–50 | 15–25 W |
| T4 Server | x86 server + RTX / T4 / A2 | Discrete GPU | TensorRT | 60–150 | 70 W+ |
| T-CPU | Any x86_64, no accelerator | None | ONNX Runtime | 2–4 | Varies |
Camera counts are per machine at a useful detection rate, and are confirmed against your own streams during the pilot.
How many cameras do you have?
Indicative edition and hardware tier for that camera count.
Deployed in a day, tuned in two weeks
No cabling, no camera procurement, no integration project. Software installs in under an hour. The rest of day one is zones and rules.
Site review
We review your camera list, resolutions, recorder and network. Output: hardware tier, cameras per machine, and the first rules to write.
Install
One installer on Linux, Windows or Docker. It detects the hardware, selects the runtime and model, and generates its own admin credentials.
Add cameras
ONVIF discovery finds cameras on the network. Others are added by RTSP URL. Credentials are encrypted at rest. Detection uses the sub-stream, so your recorder is untouched.
Select industry
Choose one of 18 packs. Zone templates, detection classes, alert rules and the KPI dashboard are applied. You adjust zones on the live image.
Tune
Rules are previewed against recent footage before going live. False alarms are one tap. Thresholds, dwell times and schedules are tightened over two weeks.
Handover
Users and roles, alert routing and escalation, scheduled reports, retention policy, signage templates and training on the incident workflow.
What you provide
- Existing IP cameras on RTSP, RTMP, HTTP/MJPEG or ONVIF. Any brand.
- A sub-stream enabled per camera, typically 640×360 or D1.
- One machine on the same network. We specify it; you buy it locally.
- Camera credentials and network access from that machine.
- One person who owns alerts.
Detection, rules, the interface and local alerts work with no internet connection.
Single site
One machine runs Edge and Hub together. Staff open it in a browser on the local network or over your VPN.
Multi-site
An Edge machine per site handles detection locally and survives a WAN outage. A central Hub gives one event list, dashboard and report set across locations.
Through an integrator
Partners get a fleet console across clients, remote health monitoring, bulk provisioning, white-label branding and reseller margin.
How the options differ
| Approach | Use your cameras | Rules & roles | Reports & compliance | Works offline | Pricing model |
|---|---|---|---|---|---|
| Cloud AI camera vendors | ✗ | ✓ | ✓ | ✗ | Per camera, per year, plus hardware |
| Enterprise VMS with analytics | ✓ | ✓ | ✓ | ✓ | Licence plus integrator project |
| Open-source NVR | ✓ | ✗ | ✗ | ✓ | Free, self-supported |
| Project-based analytics | ✓ | Partial | Partial | ✓ | Bespoke project quote |
| Ocelli | ✓ | ✓ | ✓ | ✓ | Per site |
Categories reflect the standard commercial model of each type of vendor.
Per site, not per camera
Adding a camera does not trigger a purchase decision. You pay for the site, up to your edition's camera limit.
Free
- Detection engine
- Zones and basic rules
- Local web interface
- Email and Telegram alerts
- All 18 industry packs
- 7-day event history
- Community support
One-time setup fee applies. See below.
Pro
- Everything in Free
- Plain-language rule engine
- WhatsApp, SMS, push, Slack, webhooks
- Event clips and sharing
- Scheduled PDF reports
- Retention and blurring controls
- Escalation and on-call routing
- Email support
Business
- Everything in Pro
- Multi-site dashboard
- RBAC and SSO
- REST API and MQTT
- Footage search
- Per-site tuning loop
- Compliance pack
- 8×5 support
Enterprise
- Everything in Business
- Departments and focal persons
- Campaign and turnaround windows
- Signed audit reports
- On-premise high availability
- Custom model training
- Integrations and SLA
- Named account manager
Free for up to 4 cameras
The same detection engine as the paid editions, on up to four cameras at one site, with no time limit. Commercial use inside your own business is permitted.
Included
- The full detection engine
- Up to 4 cameras at one site, any brand
- Zones, basic alert rules and camera schedules
- Local web interface, live view and event list
- Email and Telegram alerts
- All 18 industry packs
- 7 days of event history
- Face blurring on by default, on-premise processing
Not included
- More than 4 cameras or more than one site
- Plain-language rule engine
- WhatsApp, SMS, push, Slack and signed webhooks
- Event clips, clip sharing and scheduled reports
- Footage search and the tuning loop
- Roles, SSO, API and multi-site rollup
- Compliance pack and audit reports
- Warranty and support commitment
- Resale, hosting for third parties, or removing attribution
One-time setup fee
The software licence is free. Deploying it on your site is engineering work and carries a small one-time setup fee. It is not a subscription and not a per-camera charge.
- Hardware sizing and a check that your streams will run
- Installation and configuration, remote or on site
- Camera discovery, credentials and sub-stream setup
- Zone drawing and rule tuning against your footage
- Alert routing and a handover walkthrough
The amount is indicative and agreed with our sales team. It depends on your site, camera count, location and whether the work is remote or on site, and it is often credited against your first year on a paid edition. If your own IT team or integrator installs it, the software is still free.
What Ocelli does not do
No face recognition. Ocelli detects people and situations, not identities.
No emotion detection or biometric categorisation by any protected attribute.
No productivity surveillance. No idle-time tracking or individual performance scoring.
No silent cloud upload. On-premise by default. Cloud features are opt-in and labelled.
No unblurred storage by default. Un-blurring is a separate permission and is logged.
No indefinite retention. Retention defaults to 30 days, and deletion is real deletion.
No unlogged viewing. Every view of footage is recorded and visible to the account owner.
No hidden engine. The detection layer is auditable, and notice signage templates are included.
Common questions
Do I have to replace my cameras?
No. If a camera produces an RTSP, RTMP, HTTP/MJPEG or ONVIF stream, Ocelli can use it. Any brand, any age, mixed vendors on one site. We do not sell cameras.
Does my video leave the building?
Only if you enable a cloud feature. Detection, rules, alerts, the interface and search run on your own hardware. Outbound channels such as WhatsApp, SMS and email need connectivity, but they send an alert, not your footage.
How do you avoid false alarms?
Detections are confirmed across multiple frames rather than triggered by pixel change. Rules can require dwell time, direction, schedule and compound conditions. New rules are previewed against recent footage before being enabled, and marking an alert as a false alarm tightens the system. The target is under five alerts per site per day after two weeks of tuning.
What hardware do I need?
For most sites, one x86 mini PC on the same network as the cameras. Larger sites use a Jetson or a GPU server. The installer reports how many cameras the machine can carry, and we confirm it against your streams during the pilot. You buy the hardware locally and it stays your asset.
How long does deployment take?
A single-site install is normally completed in a day, of which the software takes under an hour. Allow two weeks of tuning before judging alert quality. Multi-site rollouts run one site at a time, cloning the first site's configuration.
What is the setup fee on the free edition?
The licence is free for up to 4 cameras at one site, with no time limit, including commercial use in your own business. The setup fee is a one-time charge for sizing, installation and tuning. It is not a subscription and not per camera. The amount is agreed with our sales team based on your site, camera count and location, and is often credited against your first year if you upgrade. If you install it yourself, the software is still free.
Can I upgrade from Free later?
Yes. It is the same software. A licence file raises the camera limit and unlocks the paid features in place. Cameras, zones, rules, users and event history are unaffected.
What happens if my licence expires?
Ocelli continues running for a 14-day grace period, so a renewal delay does not leave your site unmonitored. Your data remains yours and stays exportable.
Does it do face recognition?
No, by design. Ocelli detects people and situations, not identities, and faces are blurred before storage. This is a product policy, not a limitation we intend to remove.
Can my integrator install and support it?
Yes. Partners get a fleet console across clients, remote health monitoring, bulk provisioning, white-label branding and reseller margin. Contact us about the partner programme.
How does this help with DPDP and GDPR?
On-premise processing, face blurring by default, short retention with real deletion, access logging, camera masking and notice signage templates. Higher editions add a compliance pack with audit-ready reporting. Ocelli is a tool, not legal advice: you remain the data controller for your footage.
Which detections are available?
The core engine covers people, vehicles and common objects, combined with zones, tracking and time logic to produce intrusion, after-hours presence, loitering, line crossing, dwell, crowding, abandoned objects and vehicle events. PPE compliance, fire and smoke, falls and number-plate recognition are purpose-trained add-on packs. Confirm the packs you need with sales when scoping.
See it on your own cameras
Send us your camera list and what you need detected. You get the hardware tier, the camera count per machine, the first rules to write, and a firm price.