Images

Image analysis (vision input) and image generation. Related guides: Setup, Android fallback model, Availability, Chat.

Check getImageAnalysisAvailability() separately from text-model getAvailability() before supplying images, because text and vision availability can differ.

Image analysis

The images input and getImageAnalysisAvailability() belong to the 2.1.0 API. Match the
installed package version to the API reference; the source guide is not a
guarantee that these methods are available in an older npm release.

iOS

Image analysis uses native Foundation Models Attachment on iOS 27+ when the plugin is compiled with Xcode 27 / Swift 6.4. Each images[] item accepts either a readable local uri (content:// is Android-only) or raw Base64 / a Base64 data URL. Up to 4 images are accepted, each limited to 32 MiB after decoding. Base64 inputs are converted to bounded temporary native files and removed after generation. On iOS 27 builds, getImageAnalysisAvailability() returns the text-model status plus backend: 'foundation-models' and maxImages: 4; builds made with older Xcode report unavailable and cannot include iOS 27 vision support. Check it independently before attaching images. After a successful generation, image attachments are removed from retained chat history while the text prompt and response remain.

Android

Image analysis selects a native backend explicitly. getImageAnalysisAvailability() reports ml-kit-prompt when ML Kit Prompt's common feature status is available, because that SDK status does not expose a separate vision-capability flag. It reports litert-lm when a configured vision fallback is ready.

Android image input formats, ML Kit pixel budgets, the experimental multi-image path, imagePaths compatibility, and LiteRT-LM vision fallback configuration are documented in Android fallback model.

Image generation (iOS only)

Image generation is available on iOS 18.4+ via generateImage().

import { LocalLLM } from '@rdlabo/capacitor-local-llm';

const { pngBase64Images } = await LocalLLM.generateImage({
  prompt: 'A serene mountain lake at sunrise, photorealistic',
  count: 2,
});

const src = `data:image/png;base64,${pngBase64Images[0]}`;