Cloud Convert Workflow

Cloud Convert Workflow is the pack's most underrated node, and the one you'll want the moment you have a real workflow - LoRAs, a VAE, two-stage sampling, the works - and you'd rather not rebuild it by hand out of Cloud nodes. You paste a saved workflow JSON into it, it rewrites every local node class to its cloud equivalent, and it writes the converted graph to ComfyUI's output directory for you to load and run.

The conversion is mechanical but genuinely useful. The pack ships a class-name map that swaps CheckpointLoaderSimple → CloudCheckpointLoader, KSampler → CloudKSamplerGraph, KSamplerAdvanced → CloudKSamplerAdvanced, CLIPTextEncode → CloudCLIPTextEncode, EmptyLatentImage → CloudEmptyLatent, the LTX and Wan node families, and more. Boundary nodes - LoadImage, PreviewImage, SaveImage, VHS_VideoCombine, notes - are deliberately left alone, and the IMAGE socket type is intentionally never renamed, because that's the bridge between the local parts of your graph and the cloud subgraph.

Which format to paste

This is the one choice that bites people.

• Editor format (the default "Save" in ComfyUI): a {nodes, links, definitions} structure. Convert this and you get a fully editable graph back, node positions included - but only if your workflow has no subgraphs, because the converter doesn't expand them.
• API format ("Save (API Format)"): a flat {node_id: {class_type, inputs}} dict. Use this for anything with subgraphs - the LTX 2.0 templates are the pack's own example - because ComfyUI's frontend already flattens subgraphs into this form before it hits the converter.

If in doubt, save as API format. It's the format that just works for complex graphs.

Inputs and outputs

• workflow_json - multiline string; paste the raw JSON. Empty input raises a clear error.
• output_filename - defaults to converted_workflow.json; the result is written into ComfyUI's output directory (illegal filename characters get sanitized).

Outputs: converted_json (the converted workflow as a string) and saved_path (where it landed on disk). It's an output node, so it always runs when you queue.

How to install

cd ComfyUI/custom_nodes
git clone https://github.com/Dobidop/ComfyUI-CloudAPI-worker

Copy config.json.example to config.json, paste an API key from https://platform.comfy.org/profile/api-keys, restart. Or install via ComfyUI Manager. Note this node needs the API key too - the conversion itself is local, but the pack assumes a configured key for everything else.

Common issues

• "Could not parse workflow JSON" - you pasted something that isn't JSON, or copied the editor format when the string got truncated. Save fresh from ComfyUI and re-paste.
• Subgraph workflows come out broken - you used editor format. Re-save as API format.
• Converted graph won't run - check the console; the converter prints a summary and warnings, including which nodes it didn't know how to translate. Nodes outside the map stay local, which can leave a hole in the graph.
• PoC reality check - the pack's author is upfront that this is experimental. It handles the example workflows well; exotic node packs won't have cloud equivalents and will need manual rework.