Генерация подкаст-видео с несколькими персонажами

★ 7.7 · content

ai-podcast is a Claude Code skill that generates multi-person talking head podcast videos from scratch using the inference.sh pipeline: character creation → TTS audio per turn → avatar video per turn → final merge. It supports real humans with consistent identity via trained Phota profiles, quick single-use portraits via pruna/p-image, brand mascots and illustrated characters via Google Gemini 3 Pro, and mixed casts. Audio is synthesized with inworld/text-to-speech-2 at a recommended speaking rate of 1.05×, while pruna/p-video-avatar handles lip-synced animation — clips must be generated sequentially to avoid GPU conflicts. The skill is designed for anyone building podcasts, promotional conversation videos, demo reels, or any multi-speaker video content.