nodejs-ai-image-generator

Generate images using Ollama image models from Node.js.

Requirements

• Ollama running locally (default: http://localhost:11434)
• An image-capable model pulled, e.g. ollama pull x/flux2-klein:4b (Experimental, currently available only on macOS. Follow this repo for feature updates :) )

Installation

npm install nodejs-ai-image-generator

Usage

import { generateImage } from 'nodejs-ai-image-generator';

const result = await generateImage({
  prompt: 'a futuristic city at sunset, highly detailed, cinematic lighting',
  // optional — defaults shown below
  host: 'http://localhost:11434',
  model: 'x/flux2-klein:4b',
  options: {
    seed: 42, // Don't pass if you want to generate different images every time
    width: 512,
    height: 512,
    steps: 20,
    negative_prompt: 'blurry, low quality',
  },
});

// result.imageBase64 — base64-encoded PNG
// result.model       — model name used
// result.createdAt   — ISO timestamp

Save the image to disk:

import fs from 'fs';

fs.writeFileSync('output.png', Buffer.from(result.imageBase64, 'base64'));

generateImage config

| Field | Type | Default | Description | |---|---|---|---| | prompt | string | — | Image prompt (required) | | host | string | http://localhost:11434 | Ollama host URL | | model | string | x/flux2-klein:4b | Model to use | | options.seed | number | — | Reproducibility seed | | options.width | number | — | Output width in px | | options.height | number | — | Output height in px | | options.steps | number | — | Diffusion steps | | options.negative_prompt | string | — | Things to avoid |

Batch generation

Use batchGenerateImages to generate images for multiple prompts. Images are generated one by one in sequence.

import { batchGenerateImages } from 'nodejs-ai-image-generator';

const batch = await batchGenerateImages({
  prompts: [
    'a sunset over mountains',
    'a futuristic city at night',
  ],
  countPerPrompt: 3, // generate 3 images per prompt (default: 1)
  onProgress: (completed, total) => {
    console.log(`Progress: ${completed}/${total} images generated`);
  },
  // optional — same host, model, options as generateImage
  host: 'http://localhost:11434',
  model: 'x/flux2-klein:4b',
  options: {
    width: 512,
    height: 512,
    steps: 20,
    negative_prompt: 'blurry, low quality',
  },
});

for (const { prompt, results } of batch) {
  results.forEach((result, i) => {
    fs.writeFileSync(`output-${i}.png`, Buffer.from(result.imageBase64, 'base64'));
  });
}

Batch config

| Field | Type | Default | Description | |---|---|---|---| | prompts | string[] | — | List of image prompts (required) | | countPerPrompt | number | 1 | Number of images to generate per prompt | | onProgress | (completed: number, total: number) => void | — | Called after each image finishes | | host | string | http://localhost:11434 | Ollama host URL | | model | string | x/flux2-klein:4b | Model to use | | options | GenerateImageOptions | — | Same options as generateImage |

License

MIT