Avatar Talking Video

Turn text → a lip-synced talking-head video of an avatar, in two stages:

1. Voice (TTS). Delegates to the voice-clone skill to speak the text in the avatar's cloned voice. It reuses the avatar's already-trained voice, or trains one from --source if the avatar has none.
2. Video (lip-sync). Sends a camera-angle image (image, from the avatar-camera-angles skill) plus the generated mp3 (audio) to prunaai/p-video-avatar on Replicate. When audio is supplied it drives the speech directly (the model's own voice settings are ignored), so lip-sync follows the cloned voice exactly.

Output lands in <avatar>/generated-videos/ with a manifest.json recording the text, voice, image, audio, and video params.

Requirements