Mass Surveillance of viewing is not only a camera in the street. A handful of platforms now mediate how video is hosted, recommended, streamed, and often edited: YouTube, TikTok, Twitch, and Reels on Instagram. The same firms own the encoder defaults, the recommendation ranker, and the account graph. Creators who leave lose reach; viewers who stay feed a behavioural profile. Even when you “just watch”, the session is searched — hard selectors (who, when, from where) and soft ones (pause points, replays). Open codecs exist; automation and interop are the hard part. Bortzmeyer (2018) reminds that the Internet connects humans through machines, and that everything technical is also political — including who mediates viewing (see Sources).
This dossier presents FLOSS alternatives across that stack: federated hosting and live streaming (cf. ActivityPub), local and collaborative editing, capture and transcoding you can run, and tools that treat faces in a frame as something you can protect rather than scrape. Privacy by Default for video means publishing on infrastructure you can leave. NGI / NLnet funded parts of this stack; the pages below stay on what each project does.
Video is how news, teaching, and culture move. Waiting for a better ToS on a dominant host does not restore a channel after a ban, and it does not give a school or a newsroom an inspectable encoder. Mass Surveillance of viewing is already the default; federated hosts, local tools, and encryption where it belongs are the counter — but only if people actually run them.
Practical alternatives exist: ActivityPub video hosts, local editors, open capture boxes, and computer-vision privacy tools. They are uneven. Interop and usability still lag. They change the default from “upload to the ranker” to “publish on infrastructure you can leave”. Opaque, automated moderation on the big platforms is not going away; the EFF (2026) argues accountability must keep pace when machines decide what stays visible (see Sources).
The projects below are not a product list. They are work you can join, mirror, or self-host. Ask what is missing for your channel.
Mass Surveillance of viewing is a stack problem: catalogue, ranker, encode pipeline, and capture hardware. The projects below each hit a layer. None replaces YouTube overnight. Together they show hosting, editing, live, hardware, and privacy tooling you can run or audit — Privacy by Default instead of “upload and hope”.
| PeerTube is a federated video host: communities self-host an instance, federate over ActivityPub, and can use P2P delivery so watchers help each other rather than one CDN. It is the structural alternative to a single company’s catalogue. | |
| Popularizing PeerTube improves PeerTube for viewers and creators — accessibility, moderation helpers, export/import — and ships the first official mobile watch app for the PeerTube catalogue. | |
| PeerTube Remote Transcoding moves heavy encode jobs off the origin instance so small hosts can still offer multiple resolutions without a hyperscaler account. | |
| PeerTube Scalability lets a PeerTube host run across multiple nodes and ease moderation at institution scale — so a university or public broadcaster is not forced back to YouTube for traffic. | |
| Loops is ActivityPub short video — the Fediverse counterpart to TikTok-style clips, without one company owning the For You page. | |
| Miru is web-based multi-track editing and real-time AR that aims at the quality of commercial short-form tools while staying local-first and inspectable. | |
| Kdenlive is a mature non-linear editor (NLE). Parametrised keyframes / a dope sheet give a central place to animate effect parameters on the timeline so FLOSS editing keeps pace with Premiere/Final Cut workflows. | |
![]() |
Hyper 8 is a self-hostable, maintenance-light video publishing tool — static-ish hosting for people who need to put files on the web without running a full PeerTube instance. |
| MistServer Video-on-Demand extends an open media server toward distributed cache and transcode nodes so VoD is not automatically an AWS bill. | |
| Owncast is ActivityPub livecasting: stream from OBS to a server you run, with Fediverse reach rather than a Twitch ToS. | |
| Video Box is affordable open-hardware video-to-network (FOSDEM-style capture), so events need not rent a closed encoder appliance. | |
| Vframe uses computer vision to shield privacy in video — faces and identifying detail as something you can mask before publication. | |
| Libre-SOC Video Acceleration designs open video-acceleration instructions (with ffmpeg in mind) so decode is not stuck on proprietary GPU blobs. | |
![]() |
TISG is FPGA camera gateware for encrypted video streams — capture that does not assume an untrusted uplink. |
Browse the projects in this dossier below.
Local and collaborative editing — and capture: event boxes, encoders, and tools that treat faces in a frame as something you can protect — not scrape. Widely used FLOSS editors and capture software sit beside the projects above. The point is inspectable timelines and hardware you can leave.
Federated hosting — and publish: live streams, catalogues, and light static publishing without a single catalogue or ranker (cf. ActivityPub). Centralized platforms concentrate Mass Surveillance of viewing, takedowns, and encode cost. Automated moderation at that scale decides reach without a clear appeal path — another reason communities need hosts they can inspect and leave (see Sources). Open hosts reverse that default.
Playback and light transcode on hardware you control: local players, privacy-preserving clients, and PeerTube mobile watch work. Privacy by Default for watching means the session is not a ranking input unless you choose a silo.
Live videoconferencing without a single vendor room graph: self-hostable FLOSS stacks (Jitsi, BigBlueButton) beside closed defaults (Zoom, Teams, Google Meet). Video chat privacy adds on-device feed anonymisation — also noted on the AI dossier; this page stays the Video Meet theme.