myAiVideos
myAiVideos is an open-source, automated short-video production pipeline that takes a topic in and produces a publish-ready Chinese-language video, coordinating multiple AI models across topic discovery, script generation, visuals, audio and post-production.
myAiVideos automates short-video production end to end, from topic discovery through creative generation, visuals, audio and post-production to distribution. It uses FastAPI in Python as its orchestration core and a Spring Boot service in Java as a platform governance layer handling authentication, routing, circuit breaking, metering and observability, and is organized into seven decoupled, independently replaceable pipeline layers: topic, creative, visual, audio, post-production, distribution and optimization.
A Java gateway routes between multiple AI model providers such as DeepSeek, Qwen and GLM with Resilience4j-based circuit breaking and automatic failover. Quality is governed by prompt anchoring to keep a subject's visual identity consistent across segments, CLIP-based text-image consistency checks that reject off-prompt keyframes, and a four-tier audio-video sync auto-rescue strategy. Trace IDs propagate across the Java and Python services, with Langfuse providing call-tree observability, and usage is metered per tenant without code changes.
It runs with Docker Compose, using PostgreSQL, Redis and MinIO for storage and a Vue 3 frontend.