Released
Videobot
Asynchronous AI media creation bot for generating video and music, featuring dynamic model parameter mapping, text/image-to-video workflows, and guaranteed background delivery polling.
About the project
Videobot is a high-performance asynchronous Telegram service designed to render compute-heavy AI media content (video and music) using cutting-edge generative models.
Optimized for long-running generation tasks, the system features dynamic model schema mapping and guaranteed asset delivery mechanics even amidst network interruptions.
Key Technological Solutions: • Step-by-Step Stateful Generation Flow: Interactive user wizard powered by Finite State Machines (FSM) for selecting model tiers, uploading reference assets, and fine-tuning prompt parameters. • Multimodal Workflows (Text-to-Video & Image-to-Video): Context-aware request routing that automatically switches generation modes when a user provides a reference image. • Dynamic Model Schema Adapter: Internal translation engine that parses target model metadata, sanitizes inputs, and handles file attachments for varied external API contracts. • Guaranteed Asset Delivery Pipeline: Asynchronous status polling paired with a background Fallback Poller service to resume pending deliveries and re-try sending generated media files upon connection recovery. • Hybrid FastAPI & Messaging Core: Microservice architecture integrating a Telegram Bot engine with an embedded FastAPI REST layer for health monitoring and orchestrator integration.
Features
- Multimodal video generation (Text-to-Video and Image-to-Video) Generative music and audio track creation Dynamic input parameter adaptation based on model metadata Non-blocking task queue with background status polling Resilient Fallback Poller for guaranteed asset delivery User generation history and task state tracking Containerized deployment with embedded healthcheck endpoints