Генерация видео с шопинг-аватаром

★ 6.5 · content

agentkit-multimedia-shopping is a Claude Code skill that generates multimodal content for shopping avatar videos featuring the digital host "Xiao Sheng," built on ByteDance agentkit-samples. The skill runs four sequential stages: AI-generated character and scene images in 9:16 format (≥1080×1920), TTS voice synthesis and background music output as 16 kHz mono wav files, then assembly of five 5-second scenes into a complete 25-second vertical video. Final video rendering relies on the infinitetalk-shopping-avatar skill, which produces the InfiniteTalk-specific prompts. Python dependencies are Pillow, requests, and numpy; the workflow strictly follows an image → audio → video pipeline. It is designed for automating e-commerce product videos with a virtual AI presenter in portrait format.