A streaming media brand needed cinematic b-roll generated automatically from editorial scripts, matching a specific visual identity. We built an LTX-Video pipeline with a custom brand LoRA and a render queue.
The client's editorial team was waiting 2–3 days for outsourced b-roll. Stock footage didn't match their visual identity, and bespoke commissions blew their per-segment budget. They needed b-roll generated in minutes, on-brand, and triggered automatically when an editor uploaded a script.
Benchmarked LTX-Video 2.2, Wan 2.1, HunyuanVideo, and Mochi-1 on the client's reference shots. LTX-Video won on cinematic motion and render-time ratio.
Trained a style LoRA on 1,800 frames of the client's existing footage - colour grade, framing, depth of field. Applied at inference for consistent brand look.
An LLM step converts the editor's script into a shotlist (subject, camera move, mood, duration). Each shot becomes a prompt + LoRA call.
Redis-backed job queue distributing generation across a multi-GPU node. Adaptive batching, automatic retries, and a preview-then-approve flow for the editor.
Generated MP4s pushed straight into the client's NLE bin via API. Versioned, with prompt metadata embedded for reproducibility.
“Our editors now sketch a script and have a first-pass cut ready by lunch. The brand look is consistent because the LoRA learned it.”
- Head of Video, streaming media client (name withheld)
A 30-minute call. We'll tell you whether we can help - and if not, who can.