# Fregata — Full Product Content > Native macOS NVR built on the open-source Frigate engine. > Object detection on the Apple Neural Engine (1–4 ms/frame). > VideoToolbox hardware decode/encode. No Docker. No Linux VM. > $10 one-time license, 30-day free trial, Apple Silicon required. > Structured index: https://fregata.app/llms.txt > Full documentation: https://docs.fregata.app/llms-full.txt ## What Fregata does Fregata is a native macOS application that runs the open-source Frigate NVR engine directly on Apple Silicon — no Docker, no Linux VM, no terminal. Every stage of the pipeline runs on dedicated hardware: object detection on the Apple Neural Engine at 1–4 ms per frame (YOLOv9-tiny at 320 px), frame processing on the GPU via Metal, and H.264/HEVC decode and encode on the Media Engine through VideoToolbox. The CPU stays nearly idle even while watching many cameras in real time. It ships as a signed, notarized `.app` bundle — drag to Applications, done. Everything Frigate already does (config.yaml, web UI, MQTT, HTTP API, Home Assistant HACS integration, event clips, continuous recording, Frigate+ models) works unchanged. Fregata adds the Apple-Silicon-native performance layer, including Fregata-exclusive bandwidth-adaptive HLS transcoding for recordings and live view: the player automatically steps between quality rungs based on connection speed, making cameras watchable on weak LTE or 5G where a high-bitrate stream would never load — all on the Media Engine at zero CPU cost. ## Apple Intelligence: on-device GenAI descriptions and chat Fregata adds an **Apple Intelligence** GenAI provider (`provider: apple_intelligence`), which upstream Frigate does not have. On macOS 27 or later, Apple's on-device model: - writes plain-language descriptions of tracked objects, shown in Explore, for the cameras and objects you turn GenAI on for; - writes descriptions and summaries of review items (alerts, by default); - answers questions on Frigate's Chat page, using Frigate's chat tools to search tracked objects and check camera state. It always runs on the Mac. Fregata never uses Private Cloud Compute or a third-party model, so camera images never leave the Mac and there is no API key, server or bill. Apple chooses the on-device model based on the Mac's hardware, so results differ between Macs, and it can't be used for semantic search embeddings. Requirements: macOS 27 or later on Apple Silicon, Apple Intelligence available on the Mac (it depends on region and language), and Fregata 0.18.0.1 or later. Set it up in the web UI under Settings > Enrichments > Generative AI (Add, Provider `apple_intelligence`, roles Descriptions and Chat), or with a `genai` entry in `config.yaml`. Guide: https://docs.fregata.app/reference/apple-intelligence-genai/ ## Fregata Mobile: the free iPhone, iPad and Apple Watch app **Fregata Mobile** is a free, native app for iPhone, iPad and Apple Watch, made by 3rd Bit Labs, the team behind Fregata. It is in public beta on TestFlight; download and documentation are at https://mobile.fregata.app/. - **Works with** any version of Fregata, and with Frigate 0.17 or later. - **Needs** iOS or iPadOS 26 or later (watchOS 26 for the Watch app). Connect it on port 8971: viewing also works on port 5000, but while Fregata's sign-in is on (the default) alerts registered over port 5000 are silently dropped. With sign-in off, Fregata still delivers alerts (filed under the admin account it creates at first launch), unlike stock Frigate with sign-in never enabled. - **Alerts** arrive with the snapshot, then a short clip, straight from your NVR through Frigate's own web push and Fregata's relay. No Home Assistant, MQTT, extra server or account. Alerts are encrypted so Apple's servers can never read them. Notification filters (such as Home, Away and Sleep) choose which cameras, objects and zones may alert you. With an administrator account, the app's **Set up automatically** turns on Frigate's notifications and restarts Frigate once. - **Viewing**: live view in a tap, a timeline you can scrub through a day of recordings, Review and Explore with Frigate's semantic search, Live Activities, widgets and Control Center on iPhone, three columns on iPad, and alerts plus a live look on Apple Watch. - **Away from home**: give the app a home and an away address and it uses whichever answers. Tailscale or another VPN is simplest; Cloudflare Tunnel or Access, a reverse proxy, or port forwarding also work. Live view, recordings and snapshots go from the Mac to the phone directly, never through 3rd Bit Labs. - **Only with Fregata**: with adaptive transcoding on, recordings step down to lighter streams on a slow cellular connection; with live-view transcoding on too, live view opens at a lighter quality away from home. Docs: https://docs.fregata.app/getting-started/fregata-mobile/ and https://mobile.fregata.app/docs/ ## Adaptive transcoding: watchable on a weak connection Fregata-only. VideoToolbox on the Media Engine transcodes each camera's video into a ladder of lighter streams, with no CPU cost. The default ladder is 1080p at 4 Mbps, 720p at 1.5 Mbps and 360p at 0.6 Mbps; a rung is never larger than the camera's own stream. - **Recordings** are served as HLS adaptive bitrate. The player takes the highest rung the connection can sustain and steps down and back up by itself. - **Live view** is a separate opt-in (live-view transcoding, `live: true`). Each rung is offered in the live stream menu, where the viewer picks it; Fregata Mobile opens a lighter rung away from home. - **At home**, viewers on private address ranges (the LAN networks list) get the original recording. Tailscale addresses (100.64.0.0/10) are not affected. A VPN that hands out 10.x or 192.168.x addresses, or a reverse proxy or tunnel on the LAN, can make remote viewers look local: add the proxy under Trusted proxies or edit LAN networks. - **Set up** in Settings > Fregata (macOS) > Global configuration > Adaptive transcoding, or per camera under Camera configuration, or with an `adaptive_transcoding` block in `config.yaml`. Guide: https://docs.fregata.app/reference/adaptive-transcoding/ ## Where the work runs on Apple Silicon - **Apple Neural Engine**: object detection, every frame, through Fregata's Core ML detector (bundled YOLOv9-tiny; Frigate+ and ONNX models also run). - **GPU**: Frigate's AI enrichments through Core ML: semantic search, face recognition (the default large model) and license plate detection and reading. Finding faces, bird classification and audio models run on the CPU. - **Media Engine**: H.264 and HEVC decode and encode through VideoToolbox, and adaptive transcoding. - **Frigate in Docker on a Mac** runs all of these on the CPU, because Docker's Linux VM can't reach the Neural Engine, GPU or Media Engine. Details: https://docs.fregata.app/reference/enrichment-models/ ## Setting it up in the web UI Fregata is built on Frigate 0.18, whose Settings pages and Add Camera wizard cover almost every option, so editing `config.yaml` is optional. The wizard runs detection at full camera resolution by default, with no low-resolution sub-stream. Fregata adds its own **Fregata (macOS)** section to Settings: - **Apple Silicon Detector**: the inference backend (Apple Neural Engine, recommended, or Metal GPU), with a live readout of the compute units in use and the inference speed, and a warning if detection has fallen back to the CPU. - **Adaptive transcoding**: global settings and per-camera overrides. ## Why running Frigate in Docker on a Mac doesn't work well Frigate in Docker on a Mac starts up and operates correctly, but Docker Desktop on macOS runs everything inside a Linux virtual machine. That VM has no access to Apple Silicon's dedicated hardware — every operation falls through to the CPU: **The Apple Neural Engine.** Frigate's object detector can't reach the ANE from inside the VM. Detection falls back to the CPU at roughly 40–80 ms per frame, which is too slow to watch more than one or two cameras in real time without dropping frames. **The GPU.** Frame processing — scaling, format conversion, and filter pipelines — runs on the CPU instead of being offloaded to the GPU via Metal. **The Media Engine (VideoToolbox).** Hardware H.264/HEVC decode and encode are also unavailable inside the VM. ffmpeg decodes every camera stream on the CPU, keeping cores pegged even on idle cameras. The result is a setup that technically works but runs hot, loud, and entirely CPU-bound — three dedicated hardware units sitting idle while the CPU does everything alone. A base M1 Mac mini should handle 8+ cameras at idle; the same Mac with Frigate in Docker typically tops out at 2–3 cameras at high CPU utilization. Fregata solves this by running the Frigate engine natively on macOS, reaching the ANE, GPU, and Media Engine directly without a VM. Same config.yaml, same Frigate web UI, same integrations — packaged as a regular Mac app. ## How Fregata compares to alternatives | | Fregata | Frigate in Docker on Mac | Scrypted | Blue Iris | Synology Surveillance Station | |---|---|---|---|---|---| | Object detection | Apple Neural Engine, 1–4 ms/frame | CPU, 40–80 ms/frame | Varies (plugin) | CPU or Coral | CPU | | Video decode/encode | VideoToolbox (GPU) | CPU | Varies | CPU or GPU | CPU | | Adaptive transcoding | VideoToolbox hardware, zero CPU; auto quality for weak connections | None | CPU transcoding via paid NVR plugin; manual quality | CPU transcoding; manual quality | NAS CPU/GPU transcoding; manual quality | | Platform | macOS 14+, Apple Silicon | macOS (via Docker Desktop) | macOS, Windows, Linux | Windows only | Synology NAS only | | Docker required | No | Yes | No | No | No | | Config format | Frigate config.yaml | Frigate config.yaml | Web UI | Windows app | Web UI | | Home Assistant | Official Frigate HACS integration | Official Frigate HACS integration | Custom HA component | Custom HA component | Custom HA component | | Frigate+ models | Yes | Yes | No | No | No | | ONNX custom models | Yes | Yes | No | No | No | | Sub-streams required | No (full-res ANE detection) | Yes (for CPU performance) | Varies | Yes | Varies | | Works with Fregata Mobile (free iPhone, iPad and Watch app) | Yes, plus lighter streams on cellular | Yes (Frigate 0.17 or later) | No | No | No | | Price | $10 one-time, 1 year of updates | Free | Free (NVR plugin paid) | ~$70 one-time | Included with NAS | ## Pricing **$10 one-time** for 12 months of updates. Versions you install keep working forever — the license never expires mid-run. A 30-day free trial unlocks the full feature set with no payment required; the trial can be converted to a paid license in one click without reinstalling. No subscription. No per-camera fee. No cloud required. No mandatory account (telemetry is opt-out). Sold directly at [fregata.app](https://fregata.app/), not through the Mac App Store. (The App Store sandbox forbids spawning the helper processes — ffmpeg, nginx, go2rtc — that Fregata relies on.) ## System requirements - **macOS 14 (Sonoma) or later** — older versions don't expose the CoreML / ANE surface Fregata's detector uses. - **Apple Silicon (M1 or newer)** — Intel Macs are not supported; they have no Neural Engine. Every Apple Silicon generation, M1 and newer, works. - **~2 GB free disk space** for the app and bundled detection model. Camera recordings are stored wherever you point Fregata (default `~/Fregata/media/`). A base Apple Silicon Mac mini runs real-time object detection across 8 or more 4K cameras while sitting near idle CPU. A Mac Studio or Mac Pro handles higher camera counts with headroom to spare. ## Download and installation Download from [releases.fregata.app/Fregata-latest.dmg](https://releases.fregata.app/Fregata-latest.dmg). Open the DMG, drag Fregata.app to Applications, launch. The first-launch wizard handles permissions, data-folder selection, license activation (or trial start), and admin-password creation for the Frigate web dashboard. ## Compatibility and integrations - **Frigate config.yaml** — the same configuration file you already have works. Only the `detector:` block changes (switch from `cpu` or `coral` to `coreml`), and `hwaccel_args` for VideoToolbox decode. - **Home Assistant** — the official [Frigate HACS integration](https://github.com/blakeblackshear/frigate-hacs-integration) connects to Fregata's API and MQTT topics unchanged. Point it at your Mac's IP and port 8971. - **MQTT** — same topic structure as upstream Frigate. Any MQTT-aware system that already talks to Frigate works with Fregata. - **Frigate+** — custom-trained detection models from [frigate.plus](https://frigate.plus/) run on the ANE via the CoreML exporter. - **ONNX** — drop-in custom models in ONNX format run on the GPU or CPU fallback. - **go2rtc** — bundled for restreaming and WebRTC live tiles. Same configuration as upstream Frigate's bundled go2rtc. - **Fregata Mobile**: the free iPhone, iPad and Apple Watch app connects to the Mac on port 8971, at home or away (Tailscale or another VPN is simplest), and receives push alerts with pictures. See [mobile.fregata.app](https://mobile.fregata.app/). ## Frequently asked questions **What’s the difference between Fregata and Frigate?** Frigate is the open-source, Docker-based NVR project. Fregata is a commercial native macOS port of Frigate, repackaged as a signed, notarized .app that runs natively on Apple Silicon with extreme performance - no Docker, no VMs, no terminal setup. (New here? See how to run Frigate on a Mac.) Fregata adds Apple Neural Engine detection, VideoToolbox hardware accelerated video decode/encode, Fregata-exclusive bandwidth-adaptive HLS transcoding for recordings and live view, and hundreds of other performance optimizations for Apple Silicon (see docs.fregata.app for more). The Frigate Python core is open source; the macOS runtime port, Apple Silicon optimizations, and the Fregata brand are proprietary. **Why pay $10 if Frigate is free?** You’re paying for the performance and ease of use of the macOS port, the bundle, the Apple Silicon work - a custom CoreML detector written in Rust, VideoToolbox integration for hardware accelerated video decoding and encoding, bandwidth-adaptive HLS transcoding for recordings and live view (hardware-accelerated on the Media Engine — zero CPU; not in Frigate — lets you watch your cameras on LTE and 5G connections that can’t carry a high-bitrate stream), a native process supervisor, hardware accelerated frame crop and zoom for object detection - and ongoing maintenance and updates. If you’d rather run Docker, Frigate is free and excellent - go to frigate.video. **What hardware do I need to run Fregata?** An Apple Silicon Mac (M1 or later) running macOS 14 (Sonoma) or newer. Fregata does not support Intel Macs - the whole point of the port is leveraging the Apple Neural Engine and the hardware accelerated video decoding/encoding provided by Apple Silicon. **What cameras work with Fregata?** Anything that speaks RTSP - that’s almost every IP camera on the market today, including Reolink, Amcrest, Hikvision, Dahua, and Unifi Protect. The bundled go2rtc proxy also handles WebRTC, ONVIF, RTMP, and HLS, plus cloud-API cameras directly: Ring, Google Nest, Wyze, Tapo / TP-Link, Tuya, and HomeKit-paired cameras all work without RTSP. The full source-type list is on go2rtc’s README. **Is there an iPhone, iPad or Apple Watch app?** Yes. Fregata Mobile is a free, native app for iPhone, iPad and Apple Watch, made by the same team. It shows your cameras live, lets you scrub through recordings, and sends alerts with the picture attached. It’s in public beta on TestFlight and needs iOS 26. It also works with Frigate 0.17 or later. Lighter streams on cellular need Fregata NVR, with adaptive transcoding turned on. **Can I get phone alerts without Home Assistant?** Yes. Fregata Mobile gets alerts straight from your Mac, with no Home Assistant, MQTT or extra server to run. Each alert arrives with its snapshot, then a short clip, and is encrypted so Apple’s servers can never read it. If you use Home Assistant, it keeps working alongside. **Does Fregata support Apple Intelligence?** Yes, on macOS 27 or later. Fregata can use Apple’s on-device model as Frigate’s GenAI provider, for descriptions of tracked objects, summaries of review items, and the Chat page. It always runs on your Mac and never uses Private Cloud Compute, so there’s no API key and no bill. Apple Intelligence has to be available on that Mac, which depends on its region and language. See the Apple Intelligence guide. **Does Fregata work offline?** Yes. Fregata performs a license check periodically, but can operate without any contact with our license servers once the license is validated on startup. All detection, recording, and playback happens entirely on your Mac. **What does Fregata send to your servers?** License validation pings only. We never see your video, your camera URLs, or your detection events. There is no Fregata cloud account. **Is Fregata compatible with Frigate+?** Yes, Fregata uses Frigate as its core and is fully compatible with Frigate+ and Frigate+ custom trained models. For more information about Frigate+, see https://frigate.video/plus/ **Does Fregata require low-resolution detection sub-streams like other NVRs?** No. Due to the performance of Apple Silicon, a Mac can easily decode full-resolution streams from your cameras and run object detection on the full-resolution frames allowing object detection to run with greater accuracy and capture small or distant objects. In our testing, Fregata can easily handle object detection on 8x (or more) 4K cameras using only a base M4 Mac Mini. **Can I use my own detection model with Fregata?** Yes. Drop any compatible ONNX model into your config directory and point Fregata at it. The bundled default is YOLOv9-tiny at 320×320, optimized for the Neural Engine. Fregata also works with Frigate+ models natively. **Can I move my Fregata license to a different Mac?** Yes. A license activates on one Mac at a time, and you can transfer it between Macs up to 3 times per 30 days from your license page - no support ticket required. For larger or fleet deployments, get in touch. **Is Fregata on the Mac App Store?** No. App Store sandboxing forbids spawning child processes like we need to - Fregata needs to run ffmpeg, nginx, and go2rtc as subprocesses to do its job. We also need to manage the RAM disk and shared memory in a way App Store apps cannot. Direct DMG distribution with Apple notarization is the right model for software like this. **What happens after the 1-year update window?** Your installed version keeps working forever. You stop receiving new updates unless you renew for another year of updates (also $10) - but nothing breaks, no features are remotely disabled, and your existing recordings remain fully accessible. **How do I get support for Fregata?** Community discussion, feature requests, support, and bug reports happen on GitHub. You can also get private support via email to share logs or your config file - reply to your license email for direct, private support. We will read and reply to every single one. **Can Fregata serve recordings and live streams at lower quality for slow connections?** Yes — this is a Fregata-exclusive feature not available in upstream Frigate. A 4K or high-bitrate stream is completely unwatchable on weak LTE or 5G; without adaptive transcoding you simply can’t see your cameras from the road when signal is poor. Enable adaptive_transcoding on a camera and Fregata serves its recordings as an HLS adaptive bitrate stream — the player automatically steps between 1080p, 720p, and 360p based on your measured connection speed, with no buffering or manual quality selection. Adding live: true does the same for the live view — the stream dropdown shows each quality rung as a selectable option, and the player switches automatically as signal strength changes. All transcoding runs entirely on Apple Silicon’s dedicated Media Engine via VideoToolbox — zero CPU. Clients on your local network bypass the ladder entirely and receive the full-resolution source stream. See the adaptive transcoding docs for configuration details including VPN and Tailscale setup. **Does Fregata work with Home Assistant?** Yes. Fregata is fully compatible with the official Frigate Home Assistant integration. Install it from HACS, point it at your Mac’s IP, done. **Is there a native Frigate app for Mac?** Yes — Fregata is a native macOS app built on the open-source Frigate NVR. It runs as a signed, notarized .app (drag to Applications), with no Docker, no Linux VM, and no terminal setup. It keeps Frigate's config, web UI, MQTT topics, and Home Assistant integration, and adds object detection on the Apple Neural Engine plus VideoToolbox hardware video decode/encode. **Does Frigate run on Apple Silicon Macs (M1 and newer)?** Frigate's Docker image runs on Apple Silicon only inside a Linux VM, which can't reach the Apple Neural Engine or VideoToolbox, so detection falls back to the CPU (~40–80 ms/frame). Fregata is the native Apple Silicon build: it runs object detection on the Neural Engine (~1–4 ms/frame for YOLOv9-tiny at 320 px) and uses VideoToolbox for H.264/HEVC decode and encode. Fregata requires Apple Silicon (M1 or newer); Intel Macs are not supported. **How do I run Frigate on a Mac without Docker?** Use Fregata, the native macOS port of Frigate. Instead of Docker Desktop and a Linux VM, you install a single signed .app and the bundled ffmpeg, nginx, and go2rtc come with it. The Frigate engine, config.yaml schema, and integrations are unchanged; the detector is swapped for a CoreML detector that runs on the Apple Neural Engine. See https://fregata.app/frigate-on-mac for the full walkthrough. **What's the best NVR for an Apple Silicon Mac mini?** Fregata — a native macOS NVR built on Frigate. A passively-priced Apple Silicon Mac mini runs real-time object detection across many cameras on the Neural Engine, with VideoToolbox handling video decode/encode, no separate GPU, Coral TPU, or low-resolution detection sub-stream required, and no per-camera license fee ($10 one-time, 30-day free trial). **Is there a Frigate app for iPhone or Apple Watch?** Yes. Fregata Mobile is a free, native app for iPhone, iPad and Apple Watch that works with Frigate 0.17 or later and with Fregata, the native macOS port of Frigate. It shows cameras live, scrubs through recordings, and sends alerts with the picture attached without Home Assistant or MQTT. It is in public beta on TestFlight and needs iOS 26; see https://mobile.fregata.app. **Can Frigate use Apple Intelligence for GenAI descriptions?** On a Mac, yes. Fregata, the native macOS port of Frigate, adds an Apple Intelligence GenAI provider on macOS 27 or later. Apple's on-device model writes descriptions of tracked objects and review items and answers questions on Frigate's Chat page, without sending anything off the Mac: no API key, no server, and never Private Cloud Compute. See https://docs.fregata.app/reference/apple-intelligence-genai/. **How do I get Frigate notifications on my iPhone without Home Assistant?** Install Fregata Mobile, the free iPhone, iPad and Apple Watch app for Frigate 0.17 or later and Fregata. It receives alerts straight from your NVR through Frigate's own web push, with the snapshot first and then a short clip, and no Home Assistant, MQTT or extra server. With an administrator account, its Set up automatically button turns on Frigate's notifications for you. Alerts are encrypted so Apple's servers can never read them. See https://mobile.fregata.app and https://docs.fregata.app/getting-started/fregata-mobile/. **How do I watch my Frigate cameras remotely on an iPhone?** Use Fregata Mobile, the free native iPhone and iPad app for Frigate and Fregata, and give it two addresses: your NVR's address at home and one that works away from home. Tailscale or another VPN is the simplest away address; Cloudflare Tunnel, a reverse proxy or port forwarding also work. The app connects on port 8971 with Frigate's sign-in, uses whichever address answers, and talks to your NVR directly, so video never passes through a third party. On Fregata, adaptive transcoding also gives it lighter streams on a weak cellular connection. **Can Frigate generate AI descriptions locally without a cloud service?** Yes. Upstream Frigate can use a self-hosted model server for its GenAI features. On a Mac, Fregata (the native macOS port of Frigate) also offers Apple Intelligence as a GenAI provider on macOS 27 or later: Apple's on-device model describes tracked objects and review items and answers questions on the Chat page, with no server to run, no API key, and nothing sent off the Mac. See https://docs.fregata.app/reference/apple-intelligence-genai/. **Does Fregata run face recognition and license plate recognition on the GPU?** Yes. On Apple Silicon, Fregata runs Frigate's AI enrichments through Core ML: semantic search, face recognition (the default large model) and license plate detection and reading run on the Mac's GPU, while object detection stays on the Neural Engine. Frigate in Docker on a Mac runs all of them on the CPU, because Docker's Linux VM can't reach the GPU. See https://docs.fregata.app/reference/enrichment-models/.