Ideogram AI Ideogram V4

Ideogram V4 - the model that set the text-in-image benchmark for years - as a hosted text-to-image node. You type a normal sentence, pick a ratio, and out comes an Ideogram render with readable text in the image, something open weights only caught up to recently and still don't match at this level of polish.

There's history worth knowing here. Ideogram 4 open-weighted in June 2026, and the local community's verdict was complicated: the released checkpoints are quantized only, the license is non-commercial, and the model was trained on structured JSON captions - plain English prompts underperform badly, and a safety filter is literally trained into the weights. Running it locally means learning the JSON prompt format, hunting for the filter-bypass workflow, and holding a 24GB GPU. This node is the version that skips all of it: the hosted API takes a normal prompt and handles the rest server-side. You give up the free local generations, but you get Ideogram V4 as it was meant to be prompted.

How it works

BizyTRD cloud wrapper, straightforward. prompt (required, multiline) goes up with ratio (default 1:1), mode (default Balanced), seed, and megapixels, and an IMAGE comes back. megapixels is a float from 0.1 to 2.0 (default 1) - your resolution knob, roughly. mode is a combo; Balanced is the sane default until you know what the other modes do to your use case. seed defaults to 0 with -1 for auto, so you can reproduce a good result. api_config overrides endpoint/key per request; skip_error outputs a placeholder instead of aborting the workflow.

The category label on this node is "Self-Hosted" - a nod to the fact that Ideogram 4's weights do exist out there. That's the fork in the road: run the weights yourself (free, but JSON prompting, quantized-only, non-commercial license, refusal filter) or pay per generation and keep prompting in plain language.

Inputs and outputs

• prompt (required, multiline).
• Optional: ratio (default 1:1), mode (default Balanced), seed, megapixels (0.1–2.0, default 1), api_config, skip_error.
• Outputs: image (IMAGE), request_id, urls.

Installing BizyAirPlus

ComfyUI Manager: "BizyAirPlus", or:

cd /path/to/ComfyUI/custom_nodes
git clone https://github.com/siliconflow/BizyAirPlus.git
cd BizyAirPlus
python -m pip install -r requirements.txt

Restart, account at bizyair.ai, API key, BizyAirPlus button ON. Dependencies: bizyair-cloudberry, bizytrd - no checkpoints to download.

Common issues

Megapixels is a cost lever, not a free quality slider - 2.0 megapixels per generation adds up on a batch. If you came from the local Ideogram 4 scene expecting JSON prompting, unlearn that here; the API takes prose. Otherwise the pack standard: missing button (pip show bizyair-cloudberry), cloud errors in the console, metered pricing. And note the license difference: local Ideogram 4 is non-commercial, and the hosted API carries its own terms - worth a read before you build a product on it.