Pellar — Agentic Broadcast Graphics
Pellar is Veles Productions' agentic system for broadcast graphics — a voice-first assistant that turns a journalist's plain request into a broadcast-ready news graphic in minutes, not hours.
The bottleneck we set out to remove
In every newsroom the same graphics come back edition after edition — lower thirds, tickers, maps, charts, quotes. Each one travels the same route: a reporter asks, the request queues for a graphics designer, and the file finally lands on the edit.
Pellar doesn't replace the graphics designer or the editor — it takes the repetitive production off them, so their time goes to the work that actually needs an eye.
How it works
A reporter asks — by voice or text, like a colleague across the desk — and Pellar assembles the right template, fills in the data, and renders a broadcast-ready ProRes 4444 file with alpha, straight to the edit bin.
- Voice or text in — no technical knowledge required.
- Broadcast-ready out — ProRes 4444 with alpha, EBU R95 safe area, BT.709 legal colour.
- Brand-locked — colours, fonts and layout live in the template; the agent stays inside the guidelines.
Built on an open standard
Pellar runs on OGRAF — the open HTML graphics standard published by the EBU (European Broadcasting Union). No proprietary lock-in: templates live in an open format the whole industry can read. The agentic layer Veles adds is designed to be contributed back to the standard, not kept as a closed add-on.
Where Veles fits
Pellar is developed by Veles Productions as an NVIDIA Inception member — bringing generative AI into a real broadcast pipeline that is deterministic, standards-compliant, and shaped around how newsrooms actually work.
The team behind Pellar
