Architecting an Automated Text to Video Pipeline

Automated Multi-Language Video Pipeline: Project Snapshot What we need is: A written System Architecture Report for 1. building an automated text to video generation pipeline using avatars and 2. which converts content in English into 8 different languages before the videos are generated. • Pipeline that takes English text (via webhook/API from an existing system) → translates into 7 Indian languages → generates 4K, closed-captioned avatar videos in all 8 languages (7 local languages + English) leveraging avatars via the HeyGen API → stores outputs in the cloud with a retrieval API. • Languages: Hindi, Telugu, Marathi, Kannada, Tamil, Malayalam, Bengali (+ English passthrough). • Volume: hundreds of items/day, trending toward continuous/real-time. • Plus a simple operator dashboard: job status per language, retry/override controls, basic auth. Hard constraints • Video generation engine is NOT prescribed — freelancer must evaluate and recommend the best-fit provider against cost-per-video at our volume, no restrictive per-period generation caps, and 10–12 minute length support. Do not assume HeyGen is the rendering engine. • Whichever engine is used, official REST/API only — no browser automation. • Existing HeyGen avatars/accounts already set up per language — selected via avatar_id, not created from scratch. • Fully automatic translation — no mandatory human review step in the pipeline. Mention the translation software you suggest for usage in this step and justify why. • Async/queue-based architecture required — not synchronous request handling — given the volume. • Each of the 8 language jobs must succeed/fail independently (no single point of failure). • Out of scope: publishing/distribution to any external platform. Pipeline ends at stored, retrievable video files. • Full IP transfers to client; NDA required before credential/system access is shared. Hard gate — read carefully • Report must cover: architecture diagram, tech stack + justification, translation software to be used + justification, scaling/concurrency approach, failure/retry handling, data model, security model, deployment/ops, known limitations.