AI AVATAR STUDIO

This project is a high-performance SaaS AI Avatar Studio that allows creators to build highly dynamic, automated videos utilizing realistic digital clones, personalized voiceovers, and complex contextual editing pipelines. Powered by a "Multi-Model AI Workflow" and Background Automation, the system breaks down into four core content modules: custom AI Avatar Design (image-to-avatar generation utilizing Nano Banana/HeyGen and background queues), AI Voice Cloning (voice-print mimicking via Deepgram TTS models), AI Avatar Video Generation (script, avatar, and cloned voice orchestration mapped into deep-fake talking clips), and the flagship AI Video Agent. The Video Agent serves as an advanced programmatic editing engine; instead of outputting static talking-head footage, the Agent splits scripts into intelligent scenes, dynamically processes contextual B-rolls (via multi-tier asset choices), overlays time-synced captions, and dynamically passes assets to a Remotion composition pipeline for server-side layout compilation and final cloud export. The entire ecosystem uses a strict token-deduction system mapped seamlessly to user-specific dashboard credits and multi-tier subscriptions. Core Tech Stack Next.js 16 (App Router) & TypeScript: The cutting-edge full-stack foundation managing application routing, performance typing, and ESLint conformity. Claude Code: The primary architectural coding agent utilizing automated workspace execution to implement system configurations and code layers. Clerk + Drizzle ORM + NeonDB (PostgreSQL): A unified data-and-security matrix integration driving Row Level Security (RLS), database tables, and system endpoints off a single auth token. S3-Compatible Storage: Secure storage bucket infrastructure using cloud-signed URLs for user asset tracking, rapid browser file uploads, and high-speed MP4 downloads. Trigger.dev: The central background worker platform executing multi-stage async job queues, preventing server timeouts during continuous model operations. HeyGen API & Nano Banana Model: The core AI video/image engines responsible for realistic avatar rendering and synchronous facial animations. Deepgram AI Model: The voice processing engine providing high-quality, cost-efficient custom Voice Cloning and multilingual Text-to-Speech (TTS). Remotion: The programmatic video rendering framework handling the combination of avatar videos, multi-tier B-rolls, raw audio layers, and responsive captions into the final exported media file. Shadcn UI & Tailwind CSS: The visual framework system providing developer-focused layout design, clean responsive sidebars, and rapid component development.