# Generate images with Image Generator > Use the Nodespell Image Generator Creative Tool to create new images from a prompt, choose a model, set size and quality, and make up to four takes. Source: https://docs.nodespell.com/creative-tools/image-generator/ · Last updated: 2026-11-10 Image Generator is the Nodespell Creative Tool for creating a new image from a written prompt. You describe the image, choose a model, set its size and other options, and select **Generate**. Some models also accept reference images to guide subject or style. The result is an image take on a model node in your graph. ## When should I use Image Generator? Use Image Generator when you start from words, not from an existing picture. | If you want to | Use | | -------------------------------------------------------------------------- | ------------------------------------------------- | | Create a new image from a description | Image Generator | | Control camera, lens, shot type, color look, and lighting as named choices | [Cinematic Shot](https://docs.nodespell.com/creative-tools/cinematic-shot/) | | Change an existing image with an instruction | [Image Editor](https://docs.nodespell.com/creative-tools/image-editor/) | | Reframe an existing image from a new angle | [Variations](https://docs.nodespell.com/creative-tools/variations/) | | Turn a still into motion | [Image to Video](https://docs.nodespell.com/creative-tools/image-to-video/) | ## Generate an image 1. Open **Create** and select **Image Generator** in the tool rail. 2. Check the model card at the top. The default model is GPT Image 2.5 Flare; select the card to [choose another model](https://docs.nodespell.com/creative-tools/choose-a-model/). 3. Write your prompt. 4. If the model has a **References** section, add reference images with **Choose assets**. You can pick from your Assets or upload files. 5. Set the size and any options under **Output** and **Settings**. Open **More settings** for the rest. 6. Set **Generations** to the number of images you want, from 1 to 4. 7. Select **Generate**. On a Graph canvas, add Image Generator from the `Tab` menu and use the same controls in **Parameters**. See [Creative Tools on the canvas](https://docs.nodespell.com/creative-tools/on-the-canvas/). ## Image Generator inputs and controls Image Generator has no required media input. Only the prompt is required for most models. The controls fixed by the tool itself are: | Control | What it does | Options or default | | ----------------------------------- | ------------------------------------------------------------------- | --------------------------------------- | | Model card | Chooses the model that runs | Default GPT Image 2.5 Flare | | **Previous model** / **Next model** | Steps through the tool's models without opening the picker | Create only | | **Prompt** | Describes the image you want | Required for most models | | **References** | Reference images that guide the result, when the model accepts them | Model-dependent | | **More settings** | Every other setting the model has, with a count of changed values | Collapsed | | **Generations** | How many runs to submit at once | 1 to 4, default 1 (Create only) | | **Generate** | Validates the inputs and runs the model | Shows the credit estimate on paid plans | ## Which models work with Image Generator? Image Generator lists every available model that creates images from a prompt. The picker shows these under **Featured**: - GPT Image 2.5 Flare (default) - GPT Image 2.5 Sunburst - Nano Banana 2.1 - Grok Imagine Image 2.0 - FLUX 3 Image - Recraft V4.1 **All models** lists the rest, including Nano Banana 2, Nano Banana Pro, Seedream 5 Pro, GPT Image 2, Ideogram, Recraft, and Qwen models. Model availability can change; the picker always shows the current list. ## How do the controls change with the model? The settings below **Prompt** come from the selected model. With the default model, GPT Image 2.5 Flare, you see: | Setting | Options (GPT Image 2.5 Flare) | | ------------------------------------ | --------------------------------------------------------------------------------------------------------------- | | **Image Size** | Auto, 1024x1024, 1536x1024, 1024x1536, 2048x2048, 2048x1152, 2560x1440, 3840x2160, or Custom; default 1024x1024 | | **Custom Width** / **Custom Height** | 256 to 3840 pixels, for Custom size | | **Quality** | Low, Medium, High, XHigh, Max, or Auto; default High | | **Output Format** | PNG, JPEG, or WebP; default PNG | | **Compression** | 0 to 100, default 90 | GPT Image 2.5 Flare takes a prompt only. Models such as Nano Banana 2 and Nano Banana 2.1 add **Reference images**, **Aspect ratio**, and **Resolution**; others add settings such as a seed or quality level. When you switch model, Nodespell carries your prompt and the closest matching size to the new model and lists anything that changes. ## Where does the generated image go? Selecting **Generate** adds a model node to the current Level of your graph and runs it once per generation. Each finished image is kept as a [take](https://docs.nodespell.com/graph/takes/) on that node. The **Results** area of the composer shows the current result with **Show older result** and **Show newer result**, **View fullscreen**, and **Continue with this result**. **Continue with this result** opens a choice of **Image Upscaler** (a larger, refined version) or **Image to Video** (use the image as the starting frame). Nodespell keeps the exact image as a visible Asset node and connects it to the next tool. ## Common problems | Problem | What to do | | ------------------------------------------- | ---------------------------------------------------------------------------------------------------------------------------- | | "This value is required." under the prompt | Write a prompt before you select **Generate**. | | No **References** section | The selected model does not accept reference images. Choose a model with the **Refs** badge in the picker. | | "Some generations could not be submitted." | Part of the batch failed to start. The runs that started are kept in the node's history; select **Generate** again for more. | | "The Create action could not be completed." | Nothing ran. Check your inputs and try again. | | The estimate shows **Varies** | The cost depends on your input or output size; the run button updates as you change settings. | ## Frequently asked questions ### Which model does Image Generator use by default? Image Generator starts with GPT Image 2.5 Flare. You can switch to any model the picker lists, and your prompt carries over. ### How many images can I generate at once? In Create, set **Generations** from 1 to 4. Each generation is a separate run of the same node, and each result is kept as a take. On the canvas, each run makes one take. ### Can I use reference images with Image Generator? Yes, when the selected model accepts them. Models such as Nano Banana 2.1 show a **Reference images** control; the default model takes a prompt only. ### Does Image Generator edit an existing image? No. To change an existing image, use [Image Editor](https://docs.nodespell.com/creative-tools/image-editor/) or [Variations](https://docs.nodespell.com/creative-tools/variations/). ## Related - [Make your first image](https://docs.nodespell.com/start/first-image/) — a guided first run - [Choose a model](https://docs.nodespell.com/creative-tools/choose-a-model/) — featured models, estimates, switching - [Cinematic Shot](https://docs.nodespell.com/creative-tools/cinematic-shot/) — camera-led image generation - [Takes](https://docs.nodespell.com/graph/takes/) — browse and choose results - [Image Upscaler](https://docs.nodespell.com/creative-tools/image-upscaler/) — enlarge a generated image