openvino-genai-node/types/pipelines/image2ImagePipeline.d.ts

OpenVINO™ GenAI pipelines for using from Node.js environment

import { Tensor } from "openvino-node";
import { Image2ImagePipeline as Image2ImagePipelineWrapper } from "../addon.js";
import { ImageGenerationPerfMetrics } from "../perfMetrics.js";
import { ImageGenerationConfig, ImageGenerationCallback, Image2ImagePipelineProperties } from "../utils.js";
export type Image2ImageGenerateOptions = ImageGenerationConfig & {
    /**
     * Callback invoked after each denoising step.
     * Return `true` to stop early, `false` to continue.
     */
    callback?: ImageGenerationCallback;
};
/**
 * Pipeline for image-to-image generation.
 *
 * Generates image tensor(s) for a single prompt conditioned on an input image.
 * Use the factory method Image2ImagePipeline() to create and
 * initialize an instance.
 */
export declare class Image2ImagePipeline {
    protected readonly modelPath: string;
    protected readonly device: string;
    protected pipeline: Image2ImagePipelineWrapper | null;
    protected readonly properties: Image2ImagePipelineProperties;
    constructor(modelPath: string, device: string, properties?: Image2ImagePipelineProperties);
    /**
     * Load the pipeline. Must be called once before generate().
     */
    init(): Promise<void>;
    /**
     * Generate image tensor(s) for a single prompt conditioned on the input image.
     *
     * The input `image` must be a `u8` Tensor with batched NHWC shape `[1, H, W, 3]`
     * (single image, RGB channels-last). Unbatched `[H, W, 3]` tensors are rejected.
     *
     * Batch image generation is supported through `generationConfig.num_images_per_prompt`.
     * Returned tensor shape is `[N, H, W, 3]` for batched generation and `[1, H, W, 3]` for single image.
     */
    generate(prompt: string, image: Tensor, options?: Image2ImageGenerateOptions): Promise<Tensor>;
    /**
     * Decode a latent image into an RGB image tensor using the VAE decoder.
     *
     * Useful inside a generation `callback` to obtain the intermediate image at a denoising step.
     * Returned tensor shape is `[N, H, W, 3]` with `u8` pixels.
     */
    decode(latent: Tensor): Promise<Tensor>;
    /**
     * Get performance metrics from the latest generate() call.
     */
    getPerformanceMetrics(): ImageGenerationPerfMetrics;
    /**
     * Get current image generation config.
     */
    getGenerationConfig(): ImageGenerationConfig;
    /**
     * Update image generation config.
     */
    setGenerationConfig(config: ImageGenerationConfig): void;
}