Version: 2.0.0

Changelog

PDF to Images APIs

Available from LLMWhisperer v2.64.0 (released 2026-08-06). Earlier versions do not have these endpoints.

Three new asynchronous endpoints convert each page of a PDF to an image (PNG or JPEG at 150 DPI), packaged as a ZIP:

POST /pdf-to-images — submit a PDF (query params: format, tag, file_name, pages_to_extract, max_total_image_bytes), returns a whisper_hash.
GET /pdf-to-images-status — poll conversion status with per-page progress.
GET /pdf-to-images-retrieve — one-time download of the result ZIP (page_001.png, …). Results are retained for 2 hours.

Conversion is billed per converted page. Maximum 999 pages per document; PDF input only.

From v2.67.3 (released 2026-11-05):

Page selection: pages_to_extract takes the same syntax as /whisper. Only the selected pages are converted and billed, numbered page_001 … in the ZIP. The status API reports pages and total_page_count in the processed metadata, and a selection with no valid page fails with error_code: "invalid_page_range" without charge.
Image size limit: max_total_image_bytes caps the total size of the page images. A conversion that goes over it stops and fails with error_code: "size_limit_exceeded" and is not charged. Resubmitting the same document with the same settings fails at once, without rendering again. See size limit.

warning

This version of the API is not backward compatible with the previous version. Please refer to the API documentation for more information.

The base URL for the V2 APIs is https://llmwhisperer-api.us-central.unstract.com/api/v2

All use of whisper-hash as a parameter has been replaced with whisper_hash for consistency.

mode (str, optional): The processing mode.
mark_vertical_lines (bool, optional): Whether to reproduce vertical lines in the document.
mark_horizontal_lines (bool, optional): Whether to reproduce horizontal lines in the document.
line_splitter_strategy (str, optional): The line splitter strategy to use. An advanced option for customizing the line splitting process.
lang (str, optional): The language of the document.
tag (str, optional): A tag to associate with the document. Used for auditing and tracking purposes.
file_name (str, optional): The name of the file being processed. Used for auditing and tracking purposes.
use_webhook (str, optional): The name of the webhook to call after the document is processed.
webhook_metadata (str, optional): Metadata to send to the webhook after the document is processed.

timeout (int, optional): The timeout for API requests. There is no sync mode now. All requests are async.
force_text_processing (bool, optional): Whether to force text processing. This is feature is removed
ocr_provider (str, optional): The OCR provider to use. This is superseded by mode
processing_mode (str, optional): The processing mode. This is superseded by mode
store_metadata_for_highlighting (bool, optional): Whether to store metadata for highlighting. Feature is removed. Data still available and set back when retrieve is called

Added support for webhooks. You can now register a webhook and use it to receive the processed document.