Slapshot — Depth Map

Here's what the name doesn't tell you: this node never touches your GPU. Feed it a video and it ships the file to Slapshot's cloud, runs depth estimation over every frame on their servers, and hands you the result as a download. No model to install, no VRAM to budget, no "CUDA out of memory" at 2am. You pay for that convenience in frames, API calls, and privacy - more on all three in a second.

For a single still image you honestly don't need this. Depth Anything V2 runs happily inside ComfyUI and does a stellar job locally, for free. Depth is a solved-ish problem on stills. Video is a different animal: frame-by-frame estimation flickers, it's compute-heavy, and a temporally stable depth pass across a whole shot is exactly the job you'd rather hand to someone with a dedicated pipeline. That's the slot this node fills.

Real-world places it lands:

• a z-depth pass for compositing AI footage into a Nuke or After Effects shot
• depth-graded parallax and 2.5D push-ins on a clip
• a ControlNet depth map for re-rendering a whole video with its spatial structure locked

How it works

The mechanism is the same for every node in the Slapshot pack. The node finds your video's source file on disk (it needs a real .mp4 or .mov, not a URL), uploads it to Slapshot's storage via a presigned URL, then submits a depth_map job to autopilot.slapshot.ai. A background thread polls the job status every 60 seconds - up to 10 hours, which is a long time to leave ComfyUI running. When it finishes, the inline console says "Inference completed ✓ Check your Email" and the Download Depth Map button lights up.

The node keeps the job_id, so you can re-fetch a result any time by pasting it into the job ID field and hitting Download Previous Job Result. One catch: you can only download in the format the job was originally submitted with.

The two inputs that matter

• video - anything outputting a VIDEO socket (core ComfyUI video nodes or VideoHelperSuite work). Max 3500 frames.
• export_type - JPG gives you a per-frame depth image sequence; MOV gives you a depth map video file.

That's the whole list, and it's the whole list of outputs too. Slapshot_Depth_Map is an output node with no right-side sockets - the "output" is a file landing in your browser. That trips up people who expect to wire the depth map into a ControlNet node in the same graph. You can't; download the JPGs and load them back in if that's the plan.

Install

Easiest is ComfyUI Manager - search Slapshot, click Install, restart. Or:

cd ComfyUI/custom_nodes
git clone https://github.com/slapshot-ai/comfyui-slapshot

then restart. The README claims the only dependency is requests, which is a small lie: requirements.txt actually pulls opencv-python-headless and imageio[ffmpeg] too, used to count your video's frames before upload. All lightweight wheels, and Manager handles them.

Get an API key - there's no way around it

Sign up at app.slapshot.ai, go to Profile → INTEGRATIONS → ComfyUI, and generate a key. Then put it in ComfyUI's root .env file (recommended, wins over the alternative):

SLAPSHOT_API_KEY=your-key-here

or in custom_nodes/comfyui-slapshot/config.ini:

[API]
SLAPSHOT_API_KEY = your-key-here

Restart ComfyUI after either. No key when you queue? A modal pops up with a Get API Key button that sends you straight to the settings page.

Gotchas

• Free tier is 1000 frames total, counting everything you've already inferenced. Hit the wall and you get an error telling you to upgrade.
• 3500-frame cap per video. Longer clips get rejected up front.
• The download is only in the format you chose at submission time.
• Your video leaves the machine. It's uploaded to Slapshot's servers, full stop. Fine for footage you own; a non-starter for client work under an NDA.

One last honest note: this pack is brand new (repo went up mid-2026, essentially zero community chatter, version 0.1.10). It's a vendor-published integration, MIT-licensed, and the code is small enough to skim before you hand it your API key. That's a five-minute check worth doing for any node that calls home with a credential.