import { MultimodalMetadata } from "./metadata.js";
import { BaseMultimodalModel } from "./base.js";
import { ImageProcessingResult } from "../image.js";
export type Img2TextResult = ImageProcessingResult & {
    text: string[];
};
export declare class Img2TextModel extends BaseMultimodalModel {
    private imageEncoder?;
    private textEncoder?;
    private textDecoder?;
    constructor(metadata: MultimodalMetadata);
    init: (proxy?: boolean) => Promise<number>;
    process: (imageInputs: string | string[], textInputs: string | string[]) => Promise<Img2TextResult>;
    processStream(imageInputs: string | string[], textInputs: string | string[]): AsyncIterable<string[]>;
}