Concurrent Submit | Zhenzhen_Mju

In Midjourney's web UI, after an imagine you get a 2×2 grid and you click U1/U2/U3/U4 to upscale whichever cell you like. Comfly_Mju is that same step exposed as a node - give it a task ID and flip a U button. This Concurrent Submit twin just makes it batchable: instead of upscaling one grid at a time and waiting, you queue upscale jobs into the shared 30-worker image pool and collect the results in one pass. If you've ever sat through a long Midjourney session clicking U on every grid, you already know why this exists.

The inputs that matter

• taskId - required. The task ID of the imagine (or earlier upscale/variant) whose grid you're acting on. Wire this from the preceding Comfly_Mj step.
• U1 / U2 / U3 / U4 - four booleans, one per grid cell. Flip the one you want; you can even flip several and upscale multiple cells of the same grid in one job.
• api_key - your Zhenzhen key.
• seed / skip_error - cache-control seed and the fail-silently toggle.

The concurrent flow

Output is a task (COMFLY_IMAGE_FUTURE), not an image. Wire it into ComflyConcurrent_Image_Await (30 slots) with the rest of your batch - a row of Mju nodes, one per grid you want upscaled - and the collector returns the upscaled images in slot order with a status JSON. Because upscales are chained off a prior task, keep the taskId wiring honest: upscale the wrong task and you get a perfectly sharp version of the wrong picture.

Install and gotchas

Standard pack: T8mars/Comfyui-zhenzhen via ComfyUI Manager or git clone https://github.com/T8mars/Comfyui-zhenzhen into custom_nodes, then restart. No local models, standard deps, paid token, overseas servers (proxy from China).

The one real gotcha is chaining: Mju needs the task ID of whatever you're upscaling, and those task IDs are per-session tokens - if the upstream job expired or you restarted ComfyUI, the upscale silently targets nothing. And each upscale is a billable call, so "U1, U2, U3, U4 all at once" is four calls.