---
title: "PP-DocLayoutV3 | Tarka Utilities"
description: "Layout analysis for text, titles, tables, formulas, figures, charts, and curved pages."
url: "https://tarkahq.com/utilities/pp-doclayout-v3/"
---

Document layout

# PP-DocLayoutV3

Layout analysis for text, titles, tables, formulas, figures, charts, and curved pages.

[Sign up](https://tarkahq.com/console?auth=signup) [Utility model docs](https://tarkahq.com/docs/inference/utility-models/)

- **Status**: Live
- **Languages**: Language-independent geometry
- **License**: Apache 2.0 (PaddlePaddle)

## Model ID

| Model ID | Designed for |
| --- | --- |
| `pp-doclayout-v3` | Fast document-region detection, geometry, and reading order |

Confirm access with `GET /v1/models`, then follow the [utility-model docs](https://tarkahq.com/docs/inference/utility-models/) for the canonical v1 REST or native gRPC request contract.

## Pricing

`pp-doclayout-v3`: Free for a limited time. ~~$0.00025~~ $0.00 per successful page

 USD · catalog 2026-11-04.1. Release verified 2026-11-04. EUR uses the catalog conversion. [Live catalog](https://tarka.rest/control/v1/pricing).

## Overview

PP-DocLayoutV3 is the layout half of Tarka’s Paddle document stack. Call the `pp-doclayout-v3` alias through the Tarka extension `POST /v1/document/layout` when you need document regions but do not need the recognition pass.

Each response includes the source dimensions and ordered regions. A region has a stable request-local ID, class ID, semantic label, confidence score, axis-aligned bounding box, polygon, and zero-based reading order. Choose `box`, `quadrilateral`, `polygon`, or `auto` geometry in the request.

## When to pick it

- You already have a recognition engine and only need segmentation.
- Different region types should flow to different processors.
- Your UI needs overlays or selectable document areas.
- Latency matters more than generating the page’s text.

## When to pick PaddleOCR-VL instead

Use [PaddleOCR-VL 1.6](https://tarkahq.com/utilities/paddleocr-vl-1-6/) when you want recognized content and Markdown along with its block structure. PaddleOCR-VL already uses PP-DocLayoutV3 internally, so a normal full-page OCR request does not require a separate layout call first.

The endpoint accepts one inline base64 image or data URI and never fetches remote URLs. Use `labels` to retain only selected region types and `threshold` to control confidence filtering.