Nodes/ComfyUI/Runway Aleph2 Video to Video
ComfyUI Node Runs on cloud

Runway Aleph2 Video to Video

Edit footage by describing it, and keep the original motion

By Comfy-Org·Created 4 years ago·Updated 20 days ago· 121,575
Runway Aleph2 Video to Video
  • video
  • keyframes
  • prompt_images
  • VIDEO
prompt
seed0
public_figure_thresholdlow

Take an existing clip and change what's in it without changing how it moves. That's Aleph2 in one sentence: Runway's video-to-video editing model takes your footage, keeps the original motion and timing, and transforms the look - restyle, relight, add or remove elements, change the viewpoint - all driven by a text prompt. No masks, no segmentation, no ControlNet scaffolding. You describe the edit, and the node does it. This is the closest thing to the "just tell the video what to become" workflow the whole field has been chasing, and it's a built-in ComfyUI node.

It's an API node like the rest of the partner/video family: ships with ComfyUI core, uploads your clip to Comfy Org's backend, calls Runway's Aleph2 API through the proxy, and bills your Comfy account in credits. Runway is the closed commercial video company - the same one with the messy Stable Diffusion 1.5 origin story - and this node is how you rent its editing model without a separate Runway account.

The inputs that matter

  • video - the footage to edit. Constraints are real: 2 to 30 seconds, and at 30 fps or lower. Feed it a 60 fps clip and it rejects it. Output resolution matches the input.
  • prompt - what the output should look like, 1 to 1000 characters. This is the entire creative control surface. "Restyle as neon-noir with rain" is the vibe; it handles removal ("remove the person in the background"), relighting, and viewpoint changes.
  • keyframes - optional guidance images anchored to the input video, produced by the Aleph2 Keyframe nodes (up to 5). You tell Aleph2 "this is what the scene looks like at 3 seconds," and it edits around that.
  • prompt_images - optional guidance images anchored to the output video, from the Aleph2 Prompt Image nodes (up to 5).
  • public_figure_threshold - auto or low (default), Runway's content moderation for recognizable public figures.

Output is a single VIDEO.

Where people get burned

  • Keyframes vs prompt images is a fork, not a buffet. The node rejects using both at once. Keyframes anchor to your input footage; prompt images describe what the output should look like. Pick the one that matches your control intent.
  • Timestamps must live inside the clip. Keyframe and prompt-image positions past the input duration fail validation. Aleph2 is strict because the anchoring is strict.
  • FPS is a hard gate. The 30 fps ceiling trips up people who screen-recorded at 60 or 120 for slow-mo work. Transcode before you wire it in.
  • It's the most expensive per-second node in the Runway set. The badge reads about $0.40/second, so a 10-second edit is a real line item. Draft the prompt before you render, and use a short representative clip to test the edit direction before committing to the full take.

For restyling, object removal, relighting, and re-framing existing footage without touching motion, this is the built-in to use. It's the video-editing equivalent of image inpainting, and it genuinely feels like the future - until the bill lands.

Categorypartner/video/Runway

Inputs (6)

NameTypeDefaultDescription
promptSTRINGDescribes what should appear in the output (1-1000 characters).
videoVIDEOInput video to edit. Must be 2-30 seconds at 30 fps or lower.
seedINT00–4294967295Random seed for generation
public_figure_thresholdCOMBOlowContent moderation for recognizable public figures.
keyframesoptRUNWAY_ALEPH2_KEYFRAMEGuidance images anchored to the input video, from Aleph2 Keyframe nodes (up to 5). Use keyframes or prompt images, not both.
prompt_imagesoptRUNWAY_ALEPH2_PROMPT_IMAGEGuidance images anchored to the output video, from Aleph2 Prompt Image nodes (up to 5). Use keyframes or prompt images, not both.

Outputs (1)

NameTypeDescription
VIDEOVIDEO