Skip to content

Video Generation

TomoriBot can generate short videos from a text prompt or by animating a reference image. Use /generate video, or just ask her.

  • Text-to-video — generate a short clip from a prompt.
  • Image-to-video — animate a reference image (the first image from a referenced message becomes the starting frame).
  • Looping image-to-video — when requested through chat, supported models can reuse the starting image as the final frame.
  • Customizable aspect ratios.

Image-to-video and looping depend on the selected model’s first/last-frame capabilities. TomoriBot checks OpenRouter’s current video-model catalog before submitting a paid job and asks you to remove the image, disable looping, or select a compatible model when necessary.

Video generation uses an asynchronous polling workflow: the request is submitted, then TomoriBot polls the provider until the finished clip is ready, and posts it when done. Large clips can take a while.

  1. Configure a video model with /model video.
  2. Ensure image/media generation is permitted via /capabilities.
  3. Ask her to generate, or run /generate video.

Native video generation is available on Google, OpenRouter, and Z.ai. See the full matrix in Providers & Models.

For local video generation via ComfyUI (e.g. WAN image-to-video workflows), see Setup: ComfyUI.

For the internal generation and polling architecture, see the reference on video generation.