Nodes/civitai-comfy-nodes/grok / v2.0 / createImage
ComfyUI Node

grok / v2.0 / createImage

The loose closed model, now with a resolution knob

By civitai·Created 3 months ago·Updated 2 days ago· 44
grok / v2.0 / createImage
  • api_config
  • images
  • errors
  • workflow_id
  • raw_json
prompt
output_format
image_metadata
quantity1
resolution2k
qualitymedium
aspect_ratio1:1

Grok v2.0 is the update that gave xAI's image model the controls people kept asking for: a resolution selector (1k vs 2k) and a quality toggle, on top of the loose moderation that made Grok famous. This node wraps the v2.0 text-to-image path on Civitai's orchestration layer - same autoregressive closed model, but now you get to pick whether you want the 2K output or the cheaper 1K.

Reach for this over the v1.0 node when output size or cost matters. If your pipeline feeds into an upscaler anyway, 1K with quality: low is the budget path; if the image is going straight out the door, 2K is worth the extra Buzz. And it's still Grok, which means it's still the closed generator that refuses less than Google or OpenAI's offerings.

How it works

Standard civitai-comfy-nodes mechanics: the node submits a grok / v2.0 / createImage job to the Civitai Orchestration API, long-polls, and returns the image as a native IMAGE tensor. Grok is autoregressive (token prediction, not diffusion), so there are still no sampler knobs - the new surface is resolution, quality, and the aspect-ratio menu.

The inputs that matter

  • prompt - required.
  • resolution - 1k or 2k (default 2k).
  • quality - low or medium (default medium). Lower quality is cheaper and faster.
  • aspect_ratio - auto plus the thirteen presets (2:1, 20:9, 19.5:9, 16:9, 4:3, 3:2, 1:1, 2:3, 3:4, 9:16, 9:19.5, 9:20, 1:2), default auto this time.
  • quantity - 1 to 4 per job.
  • output_format / image_metadata - format and optional metadata.

Outputs

images → Preview/Save Image. errors, workflow_id, raw_json are diagnostics; raw_json has the full job result.

Installing and authenticating

Pack-wide install:

cd ComfyUI/custom_nodes
git clone https://github.com/civitai/civitai-comfy-nodes.git
pip install -r civitai-comfy-nodes/requirements.txt

or ComfyUI Manager → "Civitai Comfy Nodes" → Install → restart. Dependencies: requests + python-socketio[client].

Auth: Civitai Auth node → CIVITAI_API_TOKEN → stored key/OAuth → browser login. Headless:

export CIVITAI_API_TOKEN=your_token

Where people get burned

  • Resolution is billing, not just pixels. 2K costs more than 1K; quality: high isn't even offered here because the API caps at medium. If you're feeding an upscaler, save the Buzz and take 1K.
  • "Least restrictive" still has a ceiling. Grok plays looser than its peers, but it's a hosted model with hosted moderation - no bypass exists for a model with no weights.
  • No seed. Reproducibility is still off the table; save what you like immediately.
  • Your prompt goes to xAI via Civitai's orchestration. Standard API-node deal.
  • v1.0 vs v2.0: if you don't need resolution/quality, v1.0 is simpler; if you want the knobs and auto aspect, this is the one.
CategoryCivitai/Image/grok

Inputs (8)

NameTypeDefaultDescription
promptSTRING
output_formatoptCOMBO4 options: , jpeg, png, webP
image_metadataoptSTRINGExternal metadata that will be stored with the image
quantityoptINT11–4
resolutionoptCOMBO2k2 options: 1k, 2k
qualityoptCOMBOmedium2 options: low, medium
aspect_ratiooptCOMBO1:1Aspect ratio: 2:1, 20:9, 19.5:9, 16:9, 4:3, 3:2, 1:1, 2:3, 3:4, 9:16, 9:19.5, 9:20, 1:2
api_configoptCIVITAI_CONFIGOptional Civitai Auth connection; defaults to CIVITAI_API_TOKEN or stored OAuth login.

Outputs (4)

NameTypeDescription
imagesIMAGE
errorsSTRING
workflow_idSTRING
raw_jsonSTRING