AssetFlow

Overview

AssetFlow is a full-stack AI production platform that orchestrates content pipelines from creation through narration, rendering, and delivery. It manages assets across stages with automated handoffs, human-in-the-loop quality review, and full pipeline visibility. Built across three major versions, each expanding capability.

Articles can enter the pipeline two ways: written directly by a user, or generated by a RAG system grounded on a document corpus. Once an article is approved, it flows through voice narration (via multiple TTS engines), slide generation, and video rendering, with review gates at each stage. The platform also handles media asset management, YouTube publishing via OAuth2, and multi-user access control.

SlideFlow, the presentation engine integrated into AssetFlow, ingests LLM-structured content and produces templated PowerPoint decks. Those slides can then feed into the video rendering pipeline, with images and video sourced from the asset library.

Key Features

• Multi-stage pipeline orchestration with status tracking and automated handoffs
• User-authored or RAG-generated articles with approval gates
• Multi-engine voice narration (F5-TTS, Qwen3-TTS, Pocket TTS) with provider fallback
• SlideFlow: independent pipeline from article to LLM-structured JSON to templated PPTX
• Two video paths: narration-driven quick render and SlideFlow presentation render
• Human-in-the-loop review gates at each pipeline stage
• Event-driven asset ingestion from cloud storage with version awareness
• Multi-user access control with per-user workflows on shared asset libraries
• YouTube integration with OAuth2 for automated publishing

Architecture

User-Written Article →─┌
                        │→ Article →──┌→ Narration → Video Render →─┌
RAG-Generated Article →─└            │       ↓          ↓           │→ YouTube
                                     │   TTS Engines  Asset Library  │
                                     │                               │
                                     └→ SlideFlow → PPTX → Video →─└
                                              ↑
                                     Article → LLM → JSON

Articles enter the pipeline either written directly by a user or generated via RAG from a document corpus. From there, two independent pipelines branch out. The narration path sends the article through TTS voice synthesis and into video rendering, compositing images from the asset library. The SlideFlow path sends the article through an LLM to produce structured JSON, which SlideFlow converts into a templated PPTX. SlideFlow presentations can also be rendered to video. Both paths can publish to YouTube.

Sample Output

Tech Stack