HeyGem任务状态

LamHeyGemQueryNode is the "did it finish yet?" node for HeyGem lip-sync jobs submitted through its sibling, LamHeyGemNode. Give it a task_id and it asks the HeyGem server (default http://localhost:8383) for the current status, then hands you back a status string, the result, and a progress number. No model loading, no GPU work on the ComfyUI side - it's a tiny polling client.

Here's the honest framing, though: LamHeyGemNode already blocks until the job finishes, polling internally and driving a ComfyUI progress bar. So in the simple case you don't need this node at all. It earns its keep when you want to decouple things - say you've submitted a job and want the rest of your graph to move on, or you're building an API/automation workflow where you collect the task_id from the submit node's output and check back later from a different graph run.

Inputs

• task_id (required) - the id that LamHeyGemNode returned. Feed it from that node's task_id output.
• server - default http://localhost:8383, tooltip 请勿修改 ("don't modify"). Change it only if your HeyGem server lives elsewhere.

If task_id is empty, it returns ("error", "No task ID provided", 0.0) rather than exploding, which is nice for guarding.

Outputs

• status (STRING) - the server's human-readable message for the task.
• result (STRING) - the output path or URL once the job is done.
• progress (FLOAT) - 0.0 to 1.0 style progress the server reports.

Status codes follow HeyGem's scheme: still processing, done, or failed - the node surfaces the server's own msg field, so "success"/failure text you see here comes straight from HeyGem.

Install

Same story as every node in this pack - it ships inside ComfyUI_Lam:

cd ComfyUI/custom_nodes
git clone https://github.com/yanlang0123/ComfyUI_Lam

or install via ComfyUI Manager by searching ComfyUI_Lam, then run the pack's install.bat / 修改文件.bat (Linux: install.sh / 修改文件.sh) as the README directs. The critical prerequisite isn't ComfyUI at all - it's a running HeyGem service on port 8383 with its own models and GPU.

Gotchas

The most common stumble is treating this as a way to check on the task from the same run while the submit node is still polling - that doesn't help, since the submit node won't finish and return the task_id until it's done. Use it across runs or in a non-blocking setup. Also note it returns the raw server message, not a normalized "done/failed" enum, so if you're branching on it you'll want to compare against what your HeyGem server actually returns. And as with the submit node, no server on 8383 means status comes back as "error".