Playbook Depth Render Pass Sequence
A zip of depth frames, ready for structure-locked video
- Images
Playbook Depth Sequence is the per-frame sibling of Playbook Depth. Instead of one depth map, it pulls a zip of depth frames from Playbook3D's cloud and returns them as a single batched IMAGE. This is your node when you want structure-locked video: a depth ControlNet applied to every frame of an animated scene, keeping each new generation pinned to the same 3D geometry.
The pitch behind the whole Playbook pack is that the depth it hands you is real - measured by the renderer's z-buffer, not estimated from pixels like MiDaS or Depth Anything. That advantage compounds over time. A single estimated depth map has errors; a video full of them has flickering errors, and flicker is the thing that kills video retexturing. Ground-truth depth from an actual 3D scene means every frame's structure agrees with every other frame, because they all came from the same camera. That's the strongest argument for this node over a local preprocessor.
How it works
Inputs: api_key (required) and run_id (optional). That's all - no fallback image, because there's no single frame to fall back to.
The mechanism is the standard Playbook dance: exchange api_key for a user token at accounts.playbook3d.com/token-wrapper/get-tokens/<key>, request download URLs (optionally scoped by run_id), and read the depth_zip key out of the response. The node downloads the zip, extracts the sorted image files to a temp directory, decodes each into a float tensor, and stacks them into a [frames, height, width, channels] batch, returned as Images.
Frame order comes from filename sorting, so zero-padded names matter. frame_0001.png through frame_0030.png will land in order; unpadded names sort lexically and you'll see 10 before 2. If your sequence comes back shuffled, fix the naming on the render side - that's the whole ordering mechanism.
Install
Same as the rest of the pack. ComfyUI Manager → Playbook Nodes, or:
cd ComfyUI/custom_nodes
git clone https://github.com/playbook3d/playbook3d-comfyui-nodes
Restart ComfyUI. No models, no extra pip dependencies - the node uses the standard library zip/tempfile machinery plus ComfyUI's built-in image stack. You do need the Playbook3D API key from beta.playbook3d.com.
Troubleshooting
The sequence nodes fail loudly, unlike their single-frame cousins. Missing or bad key → "API Key not found or incorrect." No depth zip for the given run_id → "No depth zip found for the provided parameters." - which usually means the render hasn't finished, or you passed a run_id that has no sequence attached. The console shows the underlying error before the exception, so check that first. And since there's no default_value input, a failed fetch stops the graph dead. Keep the render status in the Playbook3D app in view before queuing, and you'll mostly avoid the wall.
Inputs (2)
| Name | Type | Default | Description |
|---|---|---|---|
| api_key | STRING | — | |
| run_idopt | STRING | — |
Outputs (1)
| Name | Type | Description |
|---|---|---|
| Images | IMAGE | — |