Veo Reshoot Video (Ace Data Cloud)

You got a shot you like and the camera is just… parked. Boring. Reshoot re-renders that clip with a chosen camera movement applied - a slow push in, a pan right, a dolly-zoom - without you re-prompting the whole thing and hoping you land on the same composition.

It's the most "director" node in the pack, and also the most limited: there is no prompt field. You get a menu and that's it.

How it works

It POSTs to /veo/reshoot with an existing video ID and a motion_type, async: true, and hands back a task ID. Veo re-renders the shot with that camera move interpreted over the existing content - which is why it's a paid generation and not an edit you can undo for free. The upside is that the subject, lighting and general look survive the camera change, which is not something you can get by re-rolling the prompt.

The whole node is two widgets. That's not an unfinished node; it's the API's whole surface for this action.

The two inputs

• video_id - an original Veo video ID. Extended results aren't accepted, so if you've already run Extend on the clip, reshoot is dead for it.

• motion_type - 15 choices, default STATIONARY. The naming rewards a read rather than a click:

  • STATIONARY, and the four STATIONARY_UP / _DOWN / _LEFT / _RIGHT - the camera doesn't travel, it tilts or pans, like a locked-off tripod head.
  • STATIONARY_DOLLY_IN_ZOOM_OUT and STATIONARY_DOLLY_OUT_ZOOM_IN - the Vertigo effect. Position holds, focal length and distance fight each other, background stretches. Worth knowing it exists; it's an odd, specific look.
  • UP, DOWN, LEFT_TO_RIGHT, RIGHT_TO_LEFT - actual travel in that direction.
  • FORWARD, BACKWARD, DOLLY_IN_ZOOM_OUT, DOLLY_OUT_ZOOM_IN - moves with the dolly-zoom pair again, unattached to a stationary camera.

Outputs are status, task_id, video_url, trace_id. A fresh submission is pending with an empty video_url. Take task_id to Veo Retrieve Task and the returned video_url to Veo Load Video.

Install and key

ComfyUI Manager → Custom Nodes → search Veo → confirm publisher acedatacloud and package Veo by Ace Data Cloud → install, restart. From source:

cd /path/to/ComfyUI/custom_nodes
git clone https://github.com/AceDataCloud/VeoComfyUI.git
cd VeoComfyUI && python -m pip install -r requirements.txt

requests, and nothing else. No model files, no GPU work on your end - that's the trade for it being a metered API. The pack needs ComfyUI 0.39.0+, and it takes the key from the environment of the running process:

export ACEDATACLOUD_API_KEY='paste-your-key-here'
python main.py

Where people get burned

Mostly by expecting a prompt. There isn't one, and there's no way to nudge the content - if the reshoot changes something you liked, that's the price of the camera move, and the fix is to try a different motion_type or accept it. Treat a reshoot as a new generation with the same content, not as an adjustment to one you've already banked.

Then the ID rule. It's a video ID, not the CDN URL and not the task_id the generation nodes print - feeding it a URL gets you a blunt error instead of a mysterious one, at least. And originals only: extend a clip and you can no longer reshoot it.

Finally, the standing hazard of every submit node in this pack. It's marked as an output node, so Queue Prompt fires it whether or not anything downstream is listening, and each fire is a paid render. Mute it while you're fiddling. If a job is stuck, rerun Retrieve against the same task ID - nothing in this pack resubmits on your behalf.