Influencer Studio

Influencer Studio

Technology

ReactViteNode.jsFastifySQLiteCloudflare R2RunPod

Year

2026

About the project

Local-first production studio for consistent AI creator content. Persona identity and worlds, role-aware product composition, campaigns, storyboards, and platform finishing in one workspace.

Influencer Studio is a local-first creative production environment for building a reusable virtual creator and turning approved references into an organized flow of campaign-ready media. It sits above the generation APIs rather than competing with them: the models stay replaceable engines, and the durable value is the creative context around them — identity, worlds, products, recipes, continuity, cost visibility and delivery in one workspace.

The problem it solves is operational, not artistic. Using image and video models directly is cheap but fragmented — you re-solve the same things every session: hold a character's face and body across shots, protect a product's silhouette, materials and label text, return to the same kitchen or gym, translate a vague idea into a model-appropriate prompt, pick an engine that supports the reference roles you need, and understand what a job will cost before spending on it. The Studio makes each of those a first-class feature instead of a habit.

Three layers carry the product. **Identity and world** covers persona records, canonical references, structured appearance traits, reusable places with versions and extra views, saved products with label close-ups, and Brand Truth records that ground claims in ad copy. **Creation studios** span Images, Product Compose with role-aware multi-reference composition, Prompt Studio with 280 structured scenes, UGC Ads, Brand Lab for character-free product and web work, Campaigns, Storyboard and Content Sets. **Production operations** handle curation, asset lineage, one-variable retry chains, continuity checks, platform finishing with safe-zone preview and disclosure, calendar planning, durable job records with cost preflight, backup, and LoRA dataset preparation through Cloudflare R2 and RunPod.

Two workflows show the shape of it. A product reference plus a persona plus a place resolves into assigned reference roles, an engine preflight with a cost estimate, a composed image, a QA pass over face, hands, product geometry and label, and then motion or a content set. A Storyboard takes timestamped beats bound to persona, product and place references, compiles the script, warnings and cost for free, and renders as one continuous generation rather than an edited timeline of clips.

Influencer Studio detail
TrainRig

Next Project

TrainRig

ARXVEL — Code that scales. Design that breathes.