Concurrent Submit | Zhenzhen_Mj

Midjourney is the closed image model that Flux effectively closed the gap on, but people still use it - the style, the character handling, the whole "it just looks good" thing it does. This node is the Concurrent Submit twin of Comfly_Mj, the pack's classic Midjourney imagine node through the Zhenzhen reseller. Same API, same parameters, but with the shared image pool underneath: fire off a batch of prompts and they queue in 30 workers instead of serializing in your graph.

The inputs that matter

• text - your prompt. The one required field that does all the work.
• text_en - an optional English override; useful if you're prompting in Chinese and want the API to get a cleaner English string.
• speed - turbo mode, fast mode, relax mode. Turbo for iteration, relax when you're not in a hurry and want the cheaper queue.
• model_version - v8.2 down through v6, plus the niji line. v8.1 is the default and honestly the sweet spot for most stuff.
• ar - aspect ratio as a string (1:1, 16:9...).
• no - negative prompt in MJ's "no" syntax.
• c / s / stylize-ish knobs - c (chaos), s (stylize), iw (image weight for references), sw/cw (style/character weight), r (repeat count up to 40).
• sv - 1–4 variation strength for follow-ups.
• oref / cref / sref - character, character-style, and style references by URL. This is how you hold a face or a look across a batch.
• positive - a positive-prompt helper slot.
• video / tile - toggles for MJ video generation and seamless tiling.
• api_key - direct key paste (main route).
• seed / skip_error - cache-control seed and the fail-silently toggle.

The concurrent flow

Output is a task (COMFLY_IMAGE_FUTURE), not an image. You wire it into ComflyConcurrent_Image_Await (30 slots) along with your other jobs, and the collector waits for everything, returns images in slot order, and gives you a status JSON. The classic move: a row of these with different prompts, or one prompt with several r repeats, all feeding a single collector.

Install and gotchas

Standard pack install - T8mars/Comfyui-zhenzhen via ComfyUI Manager or git clone https://github.com/T8mars/Comfyui-zhenzhen, restart. No local models. Paid token, overseas servers (proxy from China). Midjourney-specific: r (repeat count) multiplies your bill, so r=40 on the wrong prompt is a spicy surprise; and MJ's filter is MJ's filter - the reseller can't remove it, so don't plan around content the official API refuses.