Image Generation
Every Image Generation command in CodeFizz Editor Agent, with parameters and examples.
Loading…
Every Image Generation command in CodeFizz Editor Agent, with parameters and examples.
3 commands. Each shows its parameters and an example, and has a copyable deep link, so you (or an AI agent) can jump straight to one.
list_image_modelsList the image models OpenRouter offers right now, with the parameters each one takes
Reads OpenRouter's live image model list, so a model released today is listed today. No API key is needed. Each row has the model id to pass as --model and the names of the parameters that model accepts. Filter by text, or by a parameter a model must support (for example background for transparent output). Use describe_image_model for the allowed values of each parameter and the price.
| Parameter | Type | Required | Description |
|---|---|---|---|
| filter | string | No | Only models whose id or name contains this text |
| parameter | string | No | Only models that accept this parameter (e.g. background, seed, quality) |
| max_results | int | No | Maximum rows returned (truncation is reported) |
cfa list_image_models
cfa list_image_models --filter flux
cfa list_image_models --parameter background --max-results 50describe_image_modelShow one image model's parameters with their allowed values, and its providers and prices
Returns everything needed to call a model correctly: each parameter it accepts with its rule (a list of allowed values, a minimum and maximum, or 'any value'), whether it takes reference images and how many, and each provider's price list and provider-only passthrough parameters. No API key is needed. Read this before passing --params to generate_image.
| Parameter | Type | Required | Description |
|---|---|---|---|
| model | string | Yes | Model id from list_image_models, e.g. openai/gpt-image-2.5-flare |
cfa describe_image_model --model google/gemini-nano-banana-2.1generate_imageGenerate one image or many (several prompts, several models) through OpenRouter and save them as files
Sends image requests to OpenRouter with your own API key and saves each picture as a file. One job: --prompt and --model. Many jobs: --prompts and --models as JSON arrays (every prompt runs on every model), or --jobs as a JSON array of {prompt, model, params, name, reference_images} for full control. --params applies to every job and a job's own params win. --param key=value is the same thing without JSON quotes, for shells that strip them (Windows PowerShell 5.1). Every parameter is checked against the rules the model itself publishes BEFORE anything is sent, and one bad job stops the whole call with every problem named, so no money is spent on a request that would be refused. Jobs then run in parallel and the result keeps the input order, with the file paths and the cost of each. --reference-images takes local files or URLs for image-to-image. --import-to also imports every picture into the open Unreal project as a texture. --dry-run checks everything and sends nothing. Requests that were already sent cannot be taken back: if some jobs fail after others succeeded, the result lists both.
| Parameter | Type | Required | Description |
|---|---|---|---|
| prompt | string | No | What to draw (single job) |
| model | string | No | Model id from list_image_models (single job, or the default model for --jobs entries that name none) |
| prompts | json | No | JSON array of prompts; each runs on every model in --models (or on --model) |
| models | json | No | JSON array of model ids; every prompt runs on each one |
| jobs | json | No | JSON array of {prompt, model, params, name, reference_images} for full control of each request |
| params | json | No | JSON object of model parameters for every job, e.g. {"aspect_ratio":"16:9","seed":7}; see describe_image_model |
| param | stringlist | No | Quote-free form of --params, repeatable: --param aspect_ratio=16:9 --param background=transparent. It wins over --params for the same name |
| reference_images | stringlist | No | Local image files or URLs to guide the result (image-to-image), for jobs that give none of their own |
| name | string | No | File name stem for a single job (default: made from the prompt) |
| output_dir | string | No | Folder for the files (default: the project's Saved/CodeFizzEditorAgent/Images/<time>, or./cfa-images/<time>) |
| import_to | string | No | Content path to import each picture into as a texture, e.g. /Game/Generated (needs the editor open) |
| concurrency | int | No | Requests sent at the same time (1 to 8) |
| request_timeout | int | No | Seconds to wait for one image before that job fails |
| dry_run | bool | No | Check the key, the models and every parameter, then send nothing |
cfa generate_image --model google/gemini-nano-banana-2.1 --prompt "a wooden market stall icon, flat style" --params '{"aspect_ratio":"1:1"}'
cfa generate_image --models '["google/gemini-nano-banana-2.1","black-forest-labs/flux-3-image"]' --prompts '["a red lantern","a wooden crate"]' --dry-run
cfa generate_image --json '{"jobs":[{"name":"lantern","model":"openai/gpt-image-2.5-flare","prompt":"a red paper lantern","params":{"background":"transparent"}}],"import_to":"/Game/Generated"}'
cfa generate_image --model openai/gpt-image-2.5-flare --prompt "a red potion icon" --param background=transparent --param aspect_ratio=3:2