reve / createImage

Reve is the closed image model that keeps showing up in the "when will open source catch up to OAI, Reve and Ideogram" laments - it's one of the hosted frontier generators that has no open weights at all, which means the only way to use it inside ComfyUI is a node like this one. It runs through fal's engine on Civitai's orchestration layer. Prompt in, IMAGE out, billed in Buzz.

If you've got a working local setup and a good open model, you don't need this node - that's the honest framing. Reve's appeal is the closed-model quality bar and the fact that it exists at all; if you're chasing that specific look, or you want to compare it against your local pipeline, this is the door. It's also the minimal node in this pack: no seed, no negative prompt, no sampler knobs. It's a black box with an aspect ratio.

How it works

The civitai-comfy-nodes recipe machinery: your prompt is submitted to the Civitai Orchestration API as a reve / createImage job on the fal engine, the node long-polls until it succeeds, and the returned image is converted into a native IMAGE tensor. Because Reve is a hosted model with no weights, every parameter it exposes is what fal's API exposes - and that's deliberately little.

The inputs that matter

The list is short, which is the point:

• prompt - required, and it does the heavy lifting.
• aspect_ratio - a big menu from auto, 4:1, 3:1, 21:9, 2:1, 17:9, 16:9, 3:2, 4:3, 5:4, 1:1 (default), and the portrait set on down to 1:4. auto lets the model pick from the prompt.
• quantity - 1 to 10 images per job.
• output_format / image_metadata - JPEG/PNG/WebP choice and optional metadata stamp.

That's it. There's no seed, no negative_prompt, no width/height - the API doesn't expose them, so neither does the node. If you want variety across runs, vary quantity or the prompt.

Outputs

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

Installing and authenticating

Pack-wide install - this is one of ~160 nodes in civitai-comfy-nodes:

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 in ~/.civitai/ → browser login. Headless:

export CIVITAI_API_TOKEN=your_token

Where people get burned

• It's a hosted black box. No seed means no reproducibility - you can't recreate a lucky result, and you can't A/B against a fixed seed. That's a real workflow cost for a closed model.
• Per-image billing. Reve sits on fal's infrastructure through Civitai, and fal's whole model is pay-per-call. A quantity of 10 is ten billed images. Keep that in mind.
• Your prompt leaves the machine and goes through hosted moderation - there are no weights to uncensor, and no community bypass exists for a model you can't download. If your content needs the loosest possible filtering, this is the wrong category entirely.
• Don't confuse it with the Reve edit node (CivitaiImageGenFalReveEditImage) - that one requires an input image; this one is pure text-to-image.