Image Stage

Every ComfyTV image workflow starts the same way: drop a Generate → Image stage, type a prompt, pick a workflow, hit Run. It's the "Local SD1.5" node from the README's quick tour, and it's the front door for everything else in the pack - after this node you get a picker, then edit, erase, upscale, or push frames into a video stage. If you're new to ComfyTV, this is the first node you'll actually run.

ComfyTV is jtydhr88's canvas app (~190 stages, each its own node), and Image Stage is the flagship Generate node. The workflow picker is the headline feature: it's not hard-wired to one model. It defaults to Local SD1.5, but you can bind any ComfyUI workflow as JSON - including your own - so "which backend runs this prompt" is a dropdown, not a rewrite.

How it works

The node drives a wrapped ComfyUI image workflow. The tooltip on workflow says "Which backend image workflow to invoke when Run is clicked," and the schema is honest that it's "Placeholder for now" - the real breadth comes from binding workflows in the sidebar. The inputs:

• main_prompt - your prompt. Upstream text inputs are treated as additional context.
• resolution (default 1K) and aspect_ratio (default 1:1) - combined to compute the output (w, h). The combo values are "short side in px" tiers plus ratios.
• batch_size (1–8) - how many images per Run; the wrapped sampler runs once and emits this many samples.
• texts and images - autogrow inputs (COMFY_AUTOGROW_V3) for wiring in extra prompt fragments and reference images.

Outputs: images (the full batch, COMFYTV_IMAGES) and image (the selected_index cell, 1-indexed, chosen by clicking a thumbnail in the output grid). The auto-spawned Image Picker is usually the smarter next step than grabbing image directly.

The hidden inputs, explained once

force_run_token, project_id, parent_output_id, and custom_params (a JSON of user parameter attachments) are the internal inputs, all marked "Internal" in the schema and populated by the frontend's projectStore. They're the plumbing behind per-node Run. You never set them by hand.

Installing it and getting a model

Same pack, same steps:

cd ComfyUI/custom_nodes
git clone https://github.com/jtydhr88/ComfyTV

or ComfyUI Manager ("ComfyTV"), full backend restart, then ComfyTV → Generate.

The pack itself declares no pip deps, but the Local SD1.5 workflow needs v1-5-pruned-emaonly-fp16.safetensors in ComfyUI/models/checkpoints/ (link in the README's models page). And the whole-pack gotcha: on ComfyUI Desktop, macOS, or multiple installs, clone into the running instance by absolute path, verify custom_nodes/ComfyTV/__init__.py isn't nested, and fully restart.

Common issues

• "Missing model" - no SD1.5 checkpoint in checkpoints/. Download it, or bind a workflow for a model you already have.
• All eight outputs look identical - batch runs share the sampler; that's a seed/control_after_generate style issue, not the node. The picker is there so you can roll a few batches and keep the good ones.
• Prompt context is missing - text fragments wired into texts are additional context; if they're not showing up, check the wires, and remember main_prompt is the primary.