LRA-WSHT // PRIVATE BETA

The WSHT
standard.

WSHT — Wideband Signal Handling & Transport. A signal-flow discipline operated by an LRA, a Live Robot Agent. Every ingest path, routing decision, transition, and recovery is governed by it.

Chaos arrives at the input.
Broadcast-grade signal leaves at the output.

What's in the rack
Ingest
Forgiving input. Any source, any protocol, any codec. Malformed URLs absorbed; missing metadata reconstructed.
SRTRTMPRTPNDIHLS
Relay
Zero-transcode copy-mode fan-out. Per-destination FIFO isolation — one failing output never cascades.
-c copyCrash-proof<50ms failover
Playback & Transcode
Playback, schedule, and playlist with frame-accurate cue-up — SCTE-35, GPI, and manual triggers. GPU NVENC encodes to H.264, HEVC, AV1 · up to 4K 60p HDR. Captions and metadata preserved through every encode. Graceful loop-out, hard loop-out, or crossfade-to-live at the click of a cue.
AV1HEVCH.2644K60HDR10
Switcher
Unlimited M/E with program and preview buses. Two-step arm and take. Routing groups for simultaneous multi-destination cuts.
M/EPGM / PVWArm + Take
Graphics Engine
Real-time lower-thirds, tickers, chyrons, DSK overlays. Template-driven with live-editable data. Frame-accurate insertion on every output.
DSKLower-thirdsChyronTickers
Audio Engine
Multi-channel mixing, dynamics, delay, EBU R128 loudness compliance. Per-destination audio mapping.
EBU R128AES67Multi-channel
Agent Layer
Purpose-built agents live on the system 24/7. Pre-symptomatic detection via trend correlation. Autonomous remediation. The agent layer IS the ops team; new agents are added as new monitoring concerns emerge.
24/7Self-healingPre-symptomatic
Voice Control
Natural-language broadcast commands. "Ready 2 — Take 2." The machine understands director grammar. Keyboard, touch, and API paths remain open — voice is an addition, not a replacement.
Ready N — Take NNatural language
Show Recorder
Every director decision — camera takes, transitions, graphics calls, audio cues — recorded as frame-accurate multi-track data. Replay, edit, re-time. The show becomes a training set.
Frame-accurateMulti-trackNon-destructive edit
Destinations
Destination-agnostic. Any endpoint that speaks a supported protocol. Unlimited simultaneous outputs. The system doesn't care where you send it.
SRTRTMPNDIHLS+ growing
Hardware
Plays with the gear already in your rack. NDI-compatible mixers, PTZ cameras, encoder appliances, SDI bridges. API connections to every bridge, every appliance, every camera head. Hardware-native where it matters.
NDIPTZSDI bridges
Workflow Integrations
Alerts, controls, sources, and scheduling from the tools your operation already runs on. APIs and webhooks for anything else. The list grows.
Slack Zoom Google Webhooks + growing
LIVE BROADCAST PATTERN RECORDER

The Pattern

Every show is a multi-track pattern. Video cuts, audio fades, graphic transitions, caption inserts, switching events -- all synchronized like tracks in a sequencer. re-send.it records the entire performance.

Click anywhere in the Pattern to arm the audio engine
Tempo / BPM
124
Pattern
A1
ACID (DEFAULT)
Output
Level
80%
▌SCOPE2048 pts
▌SPECTRUMFFT 1024
▌FILTERLP
The Recorder
Every Broadcast Is a Performance. Every Performance Is Data.
re-send.it records every decision the director makes -- every camera take, every graphic call, every audio cue -- as a frame-accurate multi-track pattern. Not just the output. The process.
The show becomes a dataset. The dataset becomes intelligence. A hundred recorded shows teach the system what a great director does -- and when to do it.
CAM 1
CAM 2
CAM 3
VT Roll
Switch
AI Agent
Replay
Watch any broadcast back with the director's decisions overlaid. Every cut, every call, every automation event -- frame-accurate. The show as the director experienced it.
Edit
Change a camera take. Re-time a graphic. Try a different transition. Non-destructive editing on the recorded pattern. The show becomes a living document.
Learn
Record 100 shows. The system learns the director's timing, preferences, and reactions. Eventually: autonomous show direction, trained on your judgment.
Pattern Memory
Learned Show Templates
Recorded patterns from previous shows, stored and recalled instantly. Each template is a director's playbook — camera preferences, transition timing, graphic cues — learned from real performances.
The Continuous Improvement Loop
Human direction → pattern recorded → AI analyzes → AI suggests → human corrects → better AI.
The broadcast operation teaches its own assistant, one show at a time.
Prediction Engine — Real-Time Foresight

The Pattern
Sees The Future

The same multi-track pattern that records the show also reads its future. Three metrics trend upward. The prediction line extends 30 seconds beyond the present. It turns red. The system acts — before the failure exists. Click anywhere on the oscilloscope to inject an anomaly.
Network Health — Real-Time + Prediction
T+00:00:00
RTT (ms)
Packet Loss (%)
Jitter (ms)
Prediction Horizon
CLICK ON THE OSCILLOSCOPE TO INJECT AN ANOMALY
Interactive Sandbox

Break It. Watch It Heal.

Click to introduce anomalies into the signal path. Watch the AI agents detect the disturbance, correlate the impact, and execute corrective action — all within the prediction window.
Signal Integrity Monitor
NOMINAL
CLICK TO INJECT ANOMALY — OBSERVE DETECTION AND RECOVERY
Response Timeline

From Detection
To Recovery

The entire sequence — from anomaly detection to corrective action — completes in under 50 milliseconds. No alert fatigue. The show heals itself.
-30s
Prediction
Trend analysis detects deviation from baseline
-5s
Confirmation
Multi-agent correlation confirms anomaly trajectory
0ms
Intervention
Corrective action executes at the predicted moment
<50ms
Resolution
System restored. Viewer impact: zero frames dropped
The pattern is running
Operator Console
Currently in closed beta. Broadcast professionals only.