Mock Scene Text
Fake LLM replies for wiring up your story JSON
- images
- advanced_options
- STRING
The model dropdown lists gpt-5.5-pro, gpt-5.5, gpt-5-mini, gpt-4.1, gpt-4o - and then the node calls exactly none of them. MockSceneText is the smoke-test stand-in from the Story JSON Nodes pack (stepan-bogatorjov/comfy_scenes_json_node), a per-scene text generator that hands you back a canned string so you can build and test the rest of the pipeline before any real LLM money is spent.
If you're wiring up a story workflow - prompt goes in, story JSON comes out, a save node writes it to disk - this node lets you iterate on every wire around it with zero API cost. The real text generation lives outside this pack (these "mock" nodes are placeholders for a paid generation service); here you get the plumbing plus a way to exercise it.
How it works
Simple enough to read in one sitting. If mock_response is filled in, the node returns it verbatim - that's your fake LLM output, and it's the input that matters. Leave it empty and it builds a diagnostic string like [mock:gpt-5.5-pro] prompt=... images=2 files=0 persist_context=False, so you can see what's actually arriving at the socket. The other inputs (images, files, advanced_options) are accepted and their presence is reported in that diagnostic, but nothing else happens to them.
The inputs and output that matter
- mock_response - the string to emit. Fill this with a realistic-looking story JSON and you can test every downstream node without a model call.
- prompt - what you'd send the LLM; echoed in the mock reply.
- model - purely cosmetic here. Pick whatever the real generator will use, so the swap later is a drop-in.
- persist_context - a boolean that gets echoed back; the real API node would use it to keep conversation context across scenes. Not stored here.
- Output STRING - wire it into SaveTextPassthrough's
textinput to land it asscene_NNNN.txt, which is exactly the file StoryFinalCompile reads for subtitles.
Install
This is the whole pack, which ships no heavy dependencies of its own:
cd ComfyUI/custom_nodes
git clone https://github.com/stepan-bogatorjov/comfy_scenes_json_node
Restart ComfyUI. Or use ComfyUI Manager and search for comfy_scenes_json_node (the pack registers under the display name "Story JSON Nodes").
Troubleshooting
The one thing people trip on: this node will never produce real text. It's a mock by design - if you connected it expecting GPT output, you're using the wrong half of the pack. The model choice does nothing, and setting mock_response to something that isn't valid story JSON will faithfully save that something to disk and the compiler will faithfully ignore it. Treat it as a test fixture, not a generator.
Inputs (7)
| Name | Type | Default | Description |
|---|---|---|---|
| prompt | STRING | — | |
| mock_response | STRING | — | |
| persist_context | BOOLEAN | false | — |
| model | COMBO | gpt-5.5-pro | 6 options: gpt-5.5-pro, gpt-5.5, gpt-5-mini, gpt-4.1, gpt-4o, mock |
| imagesopt | IMAGE | — | |
| filesopt | STRING | — | |
| advanced_optionsopt | DICT | — |
Outputs (1)
| Name | Type | Description |
|---|---|---|
| STRING | STRING | — |