image-to-video

To animate a still image into a video with lip-synced voiceover, the skill routes your request to the Wan 2.7 model endpoint. You provide a validated JSON payload containing your image_url, prompt, and a custom audio_url to generate the synchronized talking-head sequence.

The best way to turn a portrait photo into a video is using intent-routed model selection via the RunComfy CLI. It automatically matches your request to the HappyHorse 1.0 I2V model to produce stable identity motion for portrait animation.

Yes, you can use a reference video and audio for multi-modal image-to-video generation by routing your request to the Seedance 2.0 Pro model. It animates your still image using image, reference video, and reference audio inputs for complex scene creation.

No, you do not need to manually select an image-to-video model for different motion types. The skill uses intent-routed model selection to automatically choose the best-matching RunComfy endpoint based on your prompt and provided media inputs.

Intent-routed model selection improves image-to-video generation by matching your creative intent to specific model capabilities. It applies documented prompting presets per route to enhance output quality and reduce wasted iterations on incorrect models.

The required JSON fields to execute an image-to-video task via the RunComfy CLI are image_url and prompt. You must also include optional audio_url or video_url fields depending on the specific model endpoint requirements for lip-sync or reference media.