05 / 05 · Channel
The asset on your storefront knows which prompt drove its conversion.
Generation tools end at "download". Channel ends at "live on the storefront with performance data flowing back." Every asset that ships carries its full lineage — which LoRA version, which prompt, which judge call. Performance data flows back into the LoRA training loop, so the model gets better at making things that actually convert.
Distribution targets
What's wired in v1
- 01Shopify Admin
Hero shots, PDP gallery, and product video pushed via the Admin API. Asset metadata carries C2PA + lineage. Auto-generates compressed responsive sizes.
- 02Meta Marketing
Ad creatives + AI-label metadata to Facebook + Instagram. Multi-format export — feed, story, reel — generated from one source asset.
- 03Klaviyo
Email-ready asset bundles with hero + product detail + lifestyle. Direct integration with Klaviyo's content library — no copy-paste between tools.
- 04TikTok Shop (Phase 4)
Roadmap. Direct video push with shop-tagged products. Voice clone consistency means brand sounds the same across creator collabs.
ROAS by lineage
The performance feedback loop
Every shipped asset writes a row to BigQuery with its complete lineage tree. When Meta and Shopify return performance data — clicks, conversions, ROAS — we join on the lineage and surface which LoRA version, which judge call, which prompt skeleton drove the dollar.
That dataset trains the LoRA refinement loop. Six months in, your workspace is making content that converts measurably better than month-one outputs — not because the model improved, but because your model improved.
Compliance
Per-channel metadata, applied automatically
Each channel has its own AI-disclosure rules. Meta wants the AI label. TikTok requires an in-video badge for fully-synthetic content. EU AI Act requires C2PA provenance. Channel applies the right metadata for the right channel before push, so your team doesn't have to think about it.
By the numbers