Nodes/Comfyui-zhenzhen/zhenzhen-luma-uni-1-v1-fal
ComfyUI Node

zhenzhen-luma-uni-1-v1-fal

Text-to-image and edit in one node, up to nine reference images

By T8mars·Created about a year ago·Updated 2 days ago· 740
zhenzhen-luma-uni-1-v1-fal
  • image
  • reference_image1
  • reference_image2
  • reference_image3
  • reference_image4
  • images
  • response
  • image_urls
promptA clean editorial image with strong composition.
modetext_to_image
image_url
reference_image_urls
api_key
aspect_ratio16:9
output_formatpng
styleauto
enable_web_searchfalse
image_waybase64
poll_interval6
max_poll_attempts600
skip_errorfalse
seed0

Luma made its name on video (Dream Machine, then Ray), so its image model "Uni 1" is the quiet sibling - but it's a good one, and this node exposes it in full: text-to-image, a _max quality tier, an edit mode (with up to nine reference images for the edit), manga style, web search for the text-to-image pass, and a proper aspect-ratio ladder from 3:1 down to 1:3. It's the pack's most complete image node, even though the model it wraps doesn't get the press.

The realistic use: an all-in-one image node that can generate and edit, with Luma's clean editorial look. The edit modes take an image plus up to nine references and follow an instruction; the text_to_image modes are straight generation with optional enable_web_search (which lets the model pull current references - a genuinely rare toggle in API image nodes).

How it works

mode picks the task: text_to_image, text_to_image_max, edit, edit_max. The _max variants are the higher-quality (and pricier) tier. For edit modes, wire image (the base) and optionally reference_image1–4 or up to nine URLs in reference_image_urls. style is auto or manga (yes, Luma Uni can do manga style). Submit, poll, download, and get back images (IMAGE), response, image_urls.

Note the aspect-ratio ladder is the unusual part: auto, 3:1, 2:1, 16:9, 3:2, 1:1, 2:3, 9:16, 1:2, 1:3. Ultra-wide and ultra-tall are both available, which most image nodes skip.

Inputs that matter (the short list)

  1. mode - text_to_image to generate, edit to modify an existing image.
  2. prompt - for edit modes this is the instruction ("change the background to a desert at sunset").
  3. image + reference_image* - the edit base and the reference set.
  4. enable_web_search - text-to-image only (the tooltip is explicit); on, it grounds the generation in search results.

Installing it

Manager → search Comfyui-zhenzhen → install, or:

cd ComfyUI/custom_nodes
git clone https://github.com/T8mars/Comfyui-zhenzhen

Restart, paste your api_key into the widget. No model downloads.

Common issues

  • enable_web_search doing nothing - it's text-to-image only, per the tooltip; it's ignored in edit modes.
  • Nine references overwhelming the edit - schema supports it, quality rarely does. Like every reference-heavy node in this pack, the strong 2–4 beat the kitchen sink.
  • _max cost creep - the max tier is noticeably pricier; iterate on the base mode, export on max.
  • The pack standard gotchas - empty key, 443/VPN-in-TUN-mode for the overseas API, vendor 500s needing a re-run, fal's pre-deducted billing, cache-only seed.

One aside: Uni 1 is a "closed model, so call it" situation - no local path exists. If Luma's clean editorial look is what you need and you're already in this pack, it's a solid mid-priced image node with a wider aspect-ratio range than anything else here.

Categoryzhenzhen/FAL

Inputs (19)

NameTypeDefaultDescription
promptSTRINGA clean editorial image with strong composition.
modeoptCOMBOtext_to_image4 options: text_to_image, text_to_image_max, edit, edit_max
imageoptIMAGE
reference_image1optIMAGE
reference_image2optIMAGE
reference_image3optIMAGE
reference_image4optIMAGE
image_urloptSTRING
reference_image_urlsoptSTRINGOptional reference image URLs, one per line. Max 9.
api_keyoptSTRING
aspect_ratiooptCOMBO16:910 options: auto, 3:1, 2:1, 16:9, 3:2, 1:1, +4
output_formatoptCOMBOpng3 options: auto, png, jpeg
styleoptCOMBOauto2 options: auto, manga
enable_web_searchoptBOOLEANfalseText-to-image modes only.
image_wayoptCOMBObase642 options: base64, image_url
poll_intervaloptINT61–60
max_poll_attemptsoptINT60010–3600Default 600*6s = 3600s timeout.
skip_erroroptBOOLEANfalse
seedoptINT00–18446744073709550000Execution seed for ComfyUI cache control. Fixed reuses the cached result; randomize/increment/decrement requests a new run. This compatibility seed is not sent to APIs that do not expose a native seed parameter.

Outputs (3)

NameTypeDescription
imagesIMAGE
responseSTRING
image_urlsSTRING