Runway Aleph2 Prompt Image

This node is the "place a picture at a moment" helper for Runway's Aleph2 video editing. You feed it an image and a position on the output timeline, and it hands a little chain object to the Runway Aleph2 Video to Video node, which uses it to steer what the edited video looks like at that exact point. Think of it as a keyframe, but instead of anchoring to your input footage, it anchors to the result - a reference for how the edit should resolve.

It's a pure data-holder node, which is the nice part. It makes no API call, spends no credits, and needs no model files. All the work happens downstream, in the Aleph2 Video-to-Video node that consumes its prompt_images output. That's also the trap: this node on its own does nothing. If you drop it in a workflow without the Aleph2 editor, it just sits there being a very polite, very inert chain of images.

The inputs that matter

Just two, really:

• image - the guidance image to place at that moment.
• position - how you want to time it. The dropdown gives you two styles: Absolute time (seconds), where you type something like 3.5 for 3.5 seconds into the clip, or Fraction of duration, where 0.0 is the start and 1.0 is the end. The second one is handy when you don't know the final length and don't care - "place this at the 70% mark."

There's an optional prompt_images input for chaining. Wire the first node's prompt_images output into the second node's prompt_images input, and so on. Runway allows up to five prompt images total, so you can build a mini storyboard of the output: "frame one of the shot, then this wide, then this close-up."

Where people get burned

The sibling node Runway Aleph2 Keyframe does the same job but anchors to the input video instead of the output. Aleph2 accepts keyframes or prompt images, not both - the Video-to-Video node raises an error if you feed it both. Pick the one that matches your mental model and stick with it.

Also, if you use absolute seconds, keep it inside the clip. Aleph2 rejects a timestamp past the end of the video with a reasonably clear error, but the node only discovers the clip's real duration after the request is underway, so it's easier to just not overshoot.

The output is a custom prompt_images object, not a video. Nothing about this node is visible until the Aleph2 editor it feeds finishes its run, which is a proper cloud task with a real wait. Since it lands with the Aleph2 family (added to core in June 2026), you need a current ComfyUI to even see it in the menu.

This is a utility node, so there's not much more to say. Reach for it when you need a specific look at a specific beat of the result rather than just describing the whole edit in the prompt. It's the difference between telling Aleph2 "make it feel like a horror film" and pointing at a frame and saying "end here."