Challenge
Agency production still pays twice: once for people who can cut, and again for cloud APIs that meter every second of speech and generation. Girard Media needed a path that stays honest (no fake stems, no silent vendor hop) and stays on owned infrastructure.
Workflow
- Ingest a talking-head take into Files (workspace assets).
- Talk to timeline — the starter pipeline: own Whisper → deterministic cut → karaoke caption burn → fleet render.
- Open the editable project (
.cr) in the studio; refine with natural-language edits when needed. - Publish proof — finished films can land in the public gallery when marked public.
Result
- Free-tier users can run Talk to timeline to a downloadable karaoke-captioned MP4 plus an editable composition, with zero premium API spend on that path (Native Wow sprint, verified 2026-07).
- The public showcase on corereflex.com is fed by real generations (
GET /api/gallery,visibility=publiconly) — not stock placeholders. - Render jobs drain on the self-hosted Remotion worker (queue has completed real MP4s on MinIO).
Evidence
Operator-owned product and production stack. Metrics above are from production health and the Native Wow through-path, not a paid-customer pilot. Customer case studies will land here as they are approved.
Live deliverable (2026-07-30 fleet smoke)
- MP4: Talk-to-timeline karaoke cut (~13.6s, 16.9 MB)
- Editor: Open composition
- Kept 13520 / 21967 ms · 5 clips · 36 karaoke words · Remotion self-hosted