Concurrent Submit | Zhenzhen_Mjv

Comfly_Mjv is Midjourney's variant-and-remix step as a node: you take a finished task and tell it to zoom out, pan left, or re-render as a subtle or creative variant. The classic MJ workflow is imagine → upscale → this node - the step where you actually push a composition further. The Concurrent Submit twin makes it batchable, so you can run a whole set of variant directions on the 30-worker image pool and collect them as one batch instead of one painful render at a time.

The inputs that matter

• taskId - required. The task ID of the image you're varying.
• zoom - the out-zoom control (the "give me more room around the subject" move).
• pan_up / pan_down / pan_left / pan_right - directional pans on the existing frame.
• costume_zoom - zoom that widens the scene (the "outpaint" feel) rather than a plain crop.
• upsample_v6_2x_creative / upsample_v6_2x_subtle - the v6 2x upscale flavors: creative re-imagines detail, subtle keeps it faithful. (Note the naming: these were designed for the v6-era pipeline, and they still work as upscale variants on later versions.)
• api_key - your Zhenzhen key.
• seed / skip_error - cache-control seed and fail-silently toggle.

The concurrent flow

Output is a task (COMFLY_IMAGE_FUTURE), not an image. The intended shape: one Mjv node per direction you want to explore - one zoomed, one panned right, one creative-upsampled - all feeding ComflyConcurrent_Image_Await (30 slots), which returns the results in slot order plus a status JSON. It turns a "try five directions" decision into a single queue.

Install and gotchas

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

The usual Midjourney chaining caveat applies: taskId must point at a live task, or the variant silently targets nothing. And every direction you queue is its own billable call - "all five directions" is five calls, which is the entire point of the batch layout but also exactly how the wallet drains.