Nodes/Comfyui-zhenzhen/Concurrent Submit | zhenzhen-midjourney-lowprice
ComfyUI Node

Concurrent Submit | zhenzhen-midjourney-lowprice

Imagine, edit, pan, video

By T8mars·Created about a year ago·Updated 2 days ago· 740
Concurrent Submit | zhenzhen-midjourney-lowprice
  • image1
  • image2
  • image3
  • image4
  • end_image
  • mask
  • api_config
  • task
operationmidjourney-imagine
prompt
speedunset
size
dimensionsunset
qualityunset
style
versionunset
seed-1
negative_prompt
stylize-1
chaos-1
weird-1
tilefalse
nijifalse
iw-1.0
cw-1
sw-1
cref
sref
dref
dw-1.0
repeat0
rawfalse
draftfalse
hdfalse
stop0
extra
task_id
index-1
custom_id
directionunset
zoom_ratio2.0
modal_moderegion
video_typevid_1.1_i2v_480
animate_modemanual
motionhigh
batch_size1
metadata_json
image_url1
image_url2
image_url3
image_url4
end_url
mask_url
skip_errorfalse

This is the biggest node in the pack, and it's worth understanding what you're looking at. Comfly_midjourney_lowprice is a single node that speaks the entire Midjourney API - imagine, blend, describe, edits, upscale, variation, reroll, zoom, pan, inpaint, modal, video, and remix - through the cheap domestic (国内版) route. The author's changelog says they verified all 16 operations against the real API. This Concurrent Submit twin doesn't change any of that; it just lets you run a whole fleet of those operations in the 30-worker image pool and collect the results in one go.

Why this is a submit node

Midjourney operations are slow and interactive: you imagine, you upscale, you vary, you pan - each step a round trip. If you're doing that single-threaded in ComfyUI you'll fall asleep at the keyboard. The Submit twin returns a task handle immediately and runs the actual operation off the graph's critical path, so you can queue many operations and let the pool work them. It won't make Midjourney faster, but it makes you faster.

The inputs that matter

The full operation surface is here - the author's trick is that switching operation shows only the parameters that operation needs:

  • operation - the switch: midjourney-imagine, blend, describe, the various edit modes, upscale, variation, reroll, zoom/pan, inpaint/modal, video, remix.
  • prompt / negative_prompt - the core text inputs.
  • image1–image4 + image_url1–4 - reference images for blend/edit/video; tooltips note a local image and a URL in the same slot shouldn't both be set.
  • end_image - optional end frame for video operations.
  • mask - a ComfyUI MASK for Modal local-region redraw (white areas get redrawn).
  • task_id - needed for follow-on operations (upscale/variation/extend a previous result).
  • dimensions / size / version / quality / stylize / chaos / weird / seed / repeat / hd - the usual Midjourney parameter alphabet. size is the aspect-ratio string (e.g. 16:9) and takes precedence over dimensions when set.
  • api_config - domestic route, so wire the zhenzhen-seedance2-low-price-api-settings node rather than pasting a bare key.
  • skip_error - fail silently vs throw.

The concurrent flow

Output is a task (COMFLY_IMAGE_FUTURE). Collect your jobs with ComflyConcurrent_Image_Await (30 slots), which returns images in slot order plus a status JSON. Typical setup: several Submit nodes, each doing a different operation or a different prompt, all feeding one collector.

Install and gotchas

Pack install as always - ComfyUI Manager → "Comfyui-zhenzhen", or git clone https://github.com/T8mars/Comfyui-zhenzhen into custom_nodes, restart. No local models. Paid token, overseas servers (proxy needed from China). Two Midjourney-specific gotchas: follow-on operations need the task_id of the previous result (keep them wired), and cost adds up fast across a wide operation chain - each step is a billable call.

Categoryzhenzhen/Seedance2 Low Price/Concurrent Submit

Inputs (53)

NameTypeDefaultDescription
operationCOMBOmidjourney-imagine16 options: midjourney-imagine, midjourney-blend, midjourney-describe, midjourney-edits, midjourney-upscale, midjourney-variation, +10
promptSTRING提示词或编辑指令,支持外部 STRING 节点。
speedCOMBOunset4 options: unset, relax, fast, turbo
sizeSTRING画面比例,例如 16:9;设置后优先于 dimensions。
dimensionsCOMBOunset4 options: unset, SQUARE, PORTRAIT, LANDSCAPE
qualityCOMBOunset5 options: unset, 0.25, 0.5, 1, 2
styleSTRING
versionCOMBOunset9 options: unset, 5, 5.1, 5.2, 6, 6.1, +3
seedINT-1-1–4294967295
negative_promptSTRING
stylizeINT-1-1–1000
chaosINT-1-1–100
weirdINT-1-1–3000
tileBOOLEANfalse
nijiBOOLEANfalse
iwFLOAT-1.0-1–3
cwINT-1-1–100
swINT-1-1–1000
crefSTRING角色参考图 URL。
srefSTRING风格参考图 URL。
drefSTRING深度参考图 URL。
dwFLOAT-1.0-1–100
repeatINT00–40
rawBOOLEANfalse
draftBOOLEANfalse
hdBOOLEANfalse
stopINT00–100
extraSTRING
task_idSTRING源 Midjourney 任务 ID,可连接上游节点。
indexINT-1-1–4图像操作用 1-4,Video 任务模式用 0-3。
custom_idSTRING可选按钮 customId。
directionCOMBOunset5 options: unset, left, right, up, down
zoom_ratioFLOAT2.01–2
modal_modeCOMBOregionregion 使用遮罩;outpaint 不发送遮罩。
video_typeCOMBOvid_1.1_i2v_4804 options: vid_1.1_i2v_480, vid_1.1_i2v_720, vid_1.1_i2v_start_end_480, vid_1.1_i2v_start_end_720
animate_modeCOMBOmanual2 options: manual, auto
motionCOMBOhigh2 options: low, high
batch_sizeCOMBO13 options: 1, 2, 4
metadata_jsonSTRING可选任务 metadata JSON 对象。
image_url1STRING公网图片 URL 或 data URL 1。
image_url2STRING公网图片 URL 或 data URL 2。
image_url3STRING公网图片 URL 或 data URL 3。
image_url4STRING公网图片 URL 或 data URL 4。
end_urlSTRING可选视频结束帧 URL。
mask_urlSTRING可选 Modal 遮罩 URL。
image1optIMAGE本地图片 1,不能与同槽 image_url 同时使用。
image2optIMAGE本地图片 2,不能与同槽 image_url 同时使用。
image3optIMAGE本地图片 3,不能与同槽 image_url 同时使用。
image4optIMAGE本地图片 4,不能与同槽 image_url 同时使用。
end_imageoptIMAGE视频可选结束帧。
maskoptMASKModal 局部重绘遮罩,ComfyUI 白色区域会被重绘。
api_configoptZHENZHEN_SEEDANCE2_CONFIG
skip_erroroptBOOLEANfalse失败时返回占位结果而不是中止工作流。

Outputs (1)

NameTypeDescription
taskCOMFLY_IMAGE_FUTURE