ai-avatar-video

To create a talking-head video, provide a portrait image URL and a voiceover audio URL, and the skill routes your inputs to the appropriate RunComfy avatar model for portrait lip-sync. It generates a video matching the audio track to the portrait.

Yes, you can generate an avatar video from a text script using the script-to-video speech-insertion route. This synthesizes speech from your script and drives the character animation without requiring a separate MP3 file.

For cinematic avatar generation, use the Seedance v2 Pro route to compose reference subject visuals and reference audio. This produces more cinematic outcomes by combining multi-modal reference inputs.

Yes, you need to use the RunComfy CLI with the correct model endpoints and JSON input schema for image_url, audio_url, and prompt fields. The CLI is required to route your request to the correct avatar model.

Model routing selects the RunComfy avatar route by checking whether you have an audio file or only text, and your desired style. It automatically chooses between OmniHuman, Wan 2-7, HappyHorse, Seedance v2 Pro, or Wan 2-2 Animate.

Yes, you can generate stylized character animation by providing a prompt and an audio_url to the Wan 2-7 route. This creates a scene driven by both your text prompt and the provided audio track.