Labnana Submit Task

If you're a normal user, you can stop reading after this sentence: use Labnana Image Generation instead, because it does everything this node does plus the waiting and the downloading. LabnanaSubmitTask is the fire-and-forget version from the Labnana/Advanced menu of ComfyUI-Labnana. It submits a generation to the Labnana API, gets a task_id back immediately, and does not wait for the image. The job keeps running on Labnana's servers whether ComfyUI is open or not.

That sounds like a downgrade until you think in batches. The real use case is pipeline orchestration: you submit a dozen jobs in one run, let them cook server-side, and fetch results later with Labnana Get Task. It also survives ComfyUI restarts - the task lives on the server, not in your graph.

How it works

It posts the same generation payload as the main node to the async endpoint, which returns {taskId, status} almost instantly. Everything else is the same shared input set you already know from the generator:

• prompt - generation or editing instructions.
• model - one of the six (Gemini 3 Pro, Gemini 3.1 Flash, GPT Image 2, Wan 2.7 Pro/standard, Seedream 5.0 Pro).
• image_size (1K/2K/4K) and aspect_ratio - same per-model constraints; invalid combos fail locally before any credits are spent.
• seed - cache-busting only, never sent to the API. Randomize it if you want each run to submit a fresh job.

Optional: system_prompt (prepended style/behavior instructions), reference_images (an IMAGE batch, sent inline as base64), and reference_image_urls (one https:///gs:// URL per line). Note there's no timeout input here - this node doesn't wait, so there's nothing to time out.

Outputs: task_id (STRING - the handle you'll use to fetch the result) and status (STRING, usually pending). It's an output node, so the task_id shows in the UI and the console prints it too. Copy it down, or wire it straight into a Get Task node later.

Install

Search Labnana in ComfyUI Manager, or:

cd ComfyUI/custom_nodes
git clone https://github.com/exoticknight/ComfyUI-Labnana.git

Restart. Dependency is just requests, and you need an API key (ls_...) configured in the Labnana API Client node or via LABNANA_API_KEY.

Gotchas

• The task_id is your only handle. Lose it and you're down to digging through Labnana List Tasks to find the job again.
• Credits are spent at submission. Fire-and-forget means you're committing before you know the result - run Labnana Estimate Credits first if you're batch-spending.
• No local failure detection here. A bad prompt or an exhausted balance errors at submit time (error code 26004 = insufficient credits, 29003 = parameter error), but a job that fails after submission only shows up when you fetch it. That's what status will tell you later.

Worth it? Only if you're building something that queues many jobs at once. For one-off generations, the one-node-everything path of Image Generation is strictly less fiddly - which is exactly why the pack's own README tells you the same thing.