Skip to content

Multimodal

llmist supports image and speech generation with automatic cost tracking.

Image Generation

const result = await client.image.generate({
  model: 'dall-e-3',
  prompt: 'A 1990s desktop computer with flying toasters as a screensaver',
  size: '1024x1024',
  quality: 'hd',
});
console.log('Image URL:', result.images[0].url);
console.log('Cost:', result.cost);

Supported Models

Model	Provider	Quality	Sizes
`dall-e-3`	OpenAI	standard, hd	1024x1024, 1024x1792, 1792x1024
`dall-e-2`	OpenAI	-	256x256, 512x512, 1024x1024
`gpt-image-1`	OpenAI	low, medium, high	1024x1024, 1024x1536, 1536x1024

Save to File

const result = await client.image.generate({
  model: 'dall-e-3',
  prompt: '8-bit pixel art of a floppy disk with legs running away',
  responseFormat: 'b64_json',
});

const buffer = Buffer.from(result.images[0].b64Json!, 'base64');
fs.writeFileSync('floppy.png', buffer);

Speech Generation (TTS)

const result = await client.speech.generate({
  model: 'tts-1-hd',
  input: 'Hello, welcome to llmist!',
  voice: 'nova',
});
fs.writeFileSync('output.mp3', Buffer.from(result.audio));
console.log('Cost:', result.cost);

Supported Models

Model	Provider	Voices
`tts-1`	OpenAI	alloy, echo, fable, onyx, nova, shimmer
`tts-1-hd`	OpenAI	alloy, echo, fable, onyx, nova, shimmer

CLI Commands

bun
npm

# Image
bunx @llmist/cli image "A Windows 95 error dialog that says 'Success'" -m dall-e-3 -o success.png

# Speech
bunx @llmist/cli speech "You've got mail!" -m tts-1 --voice nova -o aol.mp3

# Image
npx @llmist/cli image "A Windows 95 error dialog that says 'Success'" -m dall-e-3 -o success.png

# Speech
npx @llmist/cli speech "You've got mail!" -m tts-1 --voice nova -o aol.mp3

Using with Gadgets

class ScreenSaverGenerator extends Gadget({
  description: 'Generates a 90s-style screensaver image',
  schema: z.object({
    style: z.enum(['flying-toasters', '3d-pipes', 'starfield', 'maze']),
  }),
}) {
  async execute(params: this['params'], ctx?: ExecutionContext): Promise<string> {
    const result = await ctx!.llmist.image.generate({
      model: 'dall-e-3',
      prompt: `A ${params.style} screensaver in the style of Windows 95`,
    });
    // Cost is automatically tracked!
    return result.images[0]?.url ?? 'Screensaver generated';
  }
}

See Also