The broadcast production suite built for 24/7 reality TV.

It captures every camera and every mic around the clock, uses AI to transcribe, identify and summarise everything that happens in the house, and puts that intelligence directly into producers' hands — searchable players, story building, roughcuts and one-drag delivery into Avid Media Composer.

From live SDI capture to an edit-ready timeline, Reality Byts turns hundreds of hours of raw footage a day into stories your team can actually find.

Live Capture & Ingest

24/7 multi-camera SDI capture

Rock-solid, GPU-accelerated capture of every fixed camera in the house, engineered to run continuously for the full length of a shoot.

Frame-accurate house clock

Every frame of video, every word of transcript and every AI event is stamped against a single master timecode, so all sources line up perfectly across days and systems.

Live proxy generation

Broadcast feeds are encoded to lightweight, edit-friendly proxies in real time, so producers can scrub and review footage moments after it happens.

File-based ingest & watch folders

Drop EVS / MXF media into a watch folder and it is automatically proxied, timecode-indexed and folded into the same searchable timeline as the live feeds.

Field & card media in the same timeline

Card rushes and PSC shoots (exit interviews, location shoots, GoPros) are ingested into the same production day, search index and player timeline as the fixed-rig footage.

Rolling safety record

A continuous background recorder keeps a rolling capture of every SDI input, with frame-accurate clip extraction on demand for anything you need to pull retrospectively.

Multi-format & multi-show ready

Configurable stream groups, camera counts, audio track maps, timecode standards (PAL/NTSC) and per-show terminology (Housemates vs Campmates vs Contestants) adapt the whole suite to your format.

AI Story Intelligence

Real-time AI transcription

Speech-to-text runs live against every mic in the house, with word-level timestamps, so the transcript is ready while the moment is still unfolding.

Automatic speaker identification

Voice fingerprinting attributes every line to the right cast member across all cameras and mics — no manual logging.

On-camera face identification & tracking

GPU vision continuously detects, tracks and names cast members on every camera, powering presence maps, “who was in the room” views and downstream automation.

AI conversation summaries

Every conversation is condensed into producer-friendly summaries and scene cards, so story teams can review a day in minutes rather than hours.

Full-text and phrase search

Search everything anyone said across the entire shoot and jump straight to the footage — by contestant, camera, day and timecode.

Beat sheets & highlight scoring

Auto-generated beat sheets and highlight scoring surface the most story-worthy moments of each day for the edit team.

House analytics that answer producer questions

Derived views of sleep patterns, room-by-room whereabouts, mic checks, diary-room logs and contestant profiles, all built from the same trusted transcript and vision data.

Broadcast-compliance sweep

An automatic censorship pass flags swearing and compliance-sensitive language across the word-level transcripts before it reaches air.

Transcript review with self-improving accuracy

Low-confidence lines are queued for human review, and corrections feed a vocabulary loop that makes the transcription engine progressively better on your show's names and slang.

Producer Tools

Browser-based multiview player

Watch any camera, solo or quad, live in the browser over WebRTC — no client software, on the facility LAN or remotely.

Full-shoot DVR

Pause, scrub and rewind across the entire archive from the same player: live and historical footage behave as one continuous timeline.

Remote archive playback

Producers off-site play the same synced multi-camera archive over a secure connection, with automatic network fallback so playback keeps working on hostile networks.

Synced ISO audio monitoring

Flip between the broadcast mix and individual contestant mics in perfect sync while reviewing any moment.

Story builder & roughcuts

Assemble clips, quotes and scenes into stories and roughcut sequences directly from the transcript and player — the shot list is the edit's starting point.

Producer story board & timeline views

Scene cards with locations, contestants, filters, timeline and board layouts give story teams multiple ways to slice a production day.

Built-in team chat

Lightweight production chat where clips, scenes, stories and quotes can be dragged straight into the conversation — the fastest way to hand a moment to another producer or an editor.

One-page insight panel

Click any contestant or moment for context: who they're with, what was said, and links straight to the footage.

Edit Integration — Avid & NLE

AAF export with automatic relink

Stories and roughcuts export as AAF with relink metadata, so proxies reconnect automatically to the full-quality masters in the edit.

Media Composer panel plugin

Reality Byts runs inside Avid Media Composer as a panel: log in, chat with the story team, and drag a roughcut straight into a bin — no per-seat install beyond a thin manifest.

Search-to-bin, no duplicate media

Phrase search and chat clips resolve to the master and group clips already in the Avid project, delivering subclips and source-monitor loads without importing a single new file.

Automated edit-prep assistant

Scheduled media scanning, database builds and clip-to-bin automation keep Avid bins current with the latest hi-res windows, without an assist editor babysitting the process.

Field-media sync forensics

For card rushes, Reality Byts computes authoritative multi-camera sync (TC slate OCR, clapper detection, waveform correlation) so mixed broadcast/GoPro/audio-recorder shoots group correctly — the part Avid can't do on its own.

Multi-NLE export

Timeline export also targets FCPXML and EDL workflows for DaVinci Resolve and Premiere-based teams.

Broadcast Audio

Facial-recognition-driven audio automix

Camera-based presence tracking automatically assigns contestant mics to automix groups on Lawo mc² consoles (via Ember+), so the room mix follows the people — live channels are never touched.

Dante/AES67 audio capture

Per-lav ISO and mix channels are captured directly from the Dante network alongside SDI-embedded audio.

Named audio channel mapping

Contestant mics and mix channels are mapped and labelled centrally (“Colin lav”, “Kitchen boom”), and those names travel through players, transcripts and edit deliverables.

Mic truth tools

Mic-check and mic-isolation views verify that every lav is on the right person and behaving, backed by the same AI attribution the transcripts use.

Security & Content Protection

Footage never leaves the facility

All AI processing (transcription, speaker ID, vision) runs on-premises on the production's own GPU hardware. No rushes, transcripts or story data are uploaded to any cloud AI service.

Trusted authentication (Clerk)

Sign-in is handled by Clerk, an industry-trusted identity platform: invite-only access from the very first login, email-code and Google sign-in, and no passwords ever stored inside Reality Byts.

Server-side access control

Roles and feature grants are enforced on every API route, not just hidden in the interface — a user without a grant cannot reach the data, full stop.

Watermarked viewing media

Burn a configurable text or image watermark (“PROXY — NOT FOR BROADCAST”, or the show's own bug) over all proxy video, switchable per production, so leaked review media is identifiable and unusable for broadcast.

Zero-trust remote access

Remote users reach the system through Cloudflare Access with identity checks in front of every request; nothing is exposed as an open port on the internet.

Encrypted in transit

TLS on every player, dashboard and API connection, on the facility LAN as well as remotely.

Signed offline licensing

Licences are cryptographically signed files verified locally — no licence server to breach, spoof or depend on.

Protected deployments

Client installs ship as protected builds without source code, and each install is tied to its licence.

Reliability & Operations

Built on industry-standard, trusted technology

The stack is proven broadcast and internet infrastructure end to end: PostgreSQL, NVIDIA DeepStream and GStreamer for media, open-standard WebRTC for playback, Prometheus/Grafana for monitoring, Clerk for identity and Cloudflare for remote access — documented, supportable technologies your engineers already know, not a black box.

Self-healing pipeline watchdog

A dedicated watchdog monitors every stage of the capture and AI pipeline and orchestrates automatic restarts before a hiccup becomes lost footage.

Self-healing proxy archive

Missing or partial proxy media is detected on a dashboard and rebuilt automatically, with manual rebuild controls and clear pending/no-source/missing states.

Full observability stack

Bundled Prometheus, Grafana and Loki give engineering a live view of every service, metric and log line across the system.

Phone alerts within minutes

Telegram alerting notifies the on-call phone when any service goes down or degrades — verified end-to-end, not just on paper.

Offline-first licensing

Licence checks are fully offline (signed local file); a network outage in a live facility can never stop the show.

Storage planning & one-page tech handout

A storage estimator computes exactly how much space a production needs from its schedule and camera plan, and prints a one-page PDF for the tech department.

Automated storage reclamation

Sleep-hour proxy pruning and an evidence-graded hi-res cleanup tool reclaim space safely — footage where something happened is never touched.

Automatic WebRTC failover

Remote playback automatically falls back across TURN relays with no client configuration, so remote producers stay connected.

Deployment & Administration

Roadmap Highlights

Avid panel: deep edit automation

Automated multigroup creation, rolling supergroup timelines with a named 1–32 audio layup, drag from Avid back into Reality Byts, and playback of Avid sequences inside the RB player. (Panel foundation and roughcut-to-bin drag are shipping now.)

Auto-markers in Media Composer

Assist-configured rules that place colour-coded markers (swearing, compliance events, keywords) on Avid clips automatically from the live transcript and vision analysis.

Music rights rundown

Drag a sequence into the panel and get a music cue sheet (song, artist, TC in/out, durations) matched against the production's music library.

Dedicated playback node

A stateless output server that scales producer playback horizontally and keeps heavy player load fully isolated from capture.

Automatic ingest failover

Ordered source ladders (Dante → SDI-embedded → file; live SDI → file backfill) with automatic arbitration and a self-healing archive that converges to best-available after any outage.

Warm-standby capture mirror

Mirrored feeds of critical cameras on a second ingest box for takeover in seconds, not a cold start.

Conversation intelligence

Topic lifecycle tracking, contestant dossiers and a relationship graph built on the existing transcript engine.

Semantic video search

“Show me moments like this” natural-language search over the footage itself, powered by video embeddings (working in-house demo today).

Per-viewer forensic watermarking

Distinct watermark policies for local vs remote sessions, with per-user identifying overlays so any leaked screen capture traces back to an individual account.