EverKind AI UGC
Internal AI UGC pipeline that turns trend inputs into persona-aware scripts, generated short-form videos, review queues, and social publishing flows.
What the product covers.
EverKind is designed around a single loop: trending topic in, reviewed video out. The pipeline analyzes 7-10 reference videos, generates scripts with persona and product context, creates AI video variants, routes them through human review, and publishes to TikTok, Instagram, and Facebook with analytics tied back to the original trend and references.
The decisions inside the build.
A concise account of the operating problem, the resulting system shape, and the technical decision that kept the work coherent.
Problem
Trend research, persona writing, media generation, editorial review, and channel publishing become unreliable when treated as one opaque AI action.
System shape
An asynchronous workflow separates reference analysis, persona and script generation, voice and video creation, review queues, publishing, and feedback signals into explicit stages.
Decision record
Slow model and media work runs in background jobs, while human review remains a required step before distribution.
What the contribution actually covered.
This is an internal product record. Public copy describes the workflow without exposing private prompts, providers, content, or channel data.
- 01
Workflow design from trend input through generation, review, and publishing.
- 02
Background-job boundaries for slow model and media operations.
- 03
Persona and product context carried through scripts, variants, and review state.
The working parts, without the theatre.
- 01Next.js
- 02BullMQ
- 03AI Video
This was a private engagement, so the public note is intentionally limited to the verified project scope.
This is an internal product record. Public copy describes the workflow without exposing private prompts, providers, content, or channel data.
No public destination published