Compare PixLab APIs for image understanding, OCR, document parsing, language models, embeddings, and structured tool calling. Find the endpoint that matches your task, then continue to its dedicated documentation for implementation details.
Use these direct routes for common tasks, or compare neighboring endpoints in the full catalog below.
Submit visual content together with a natural-language question or instruction.
query
API docs →
Convert supported documents into Markdown, JSON, or text for downstream LLM workflows.
llm-parse
API docs →
Use PixLab's OpenAI-compatible conversational endpoint for multi-turn messaging.
chat
API docs →
describe
Extract text from an image
ocr
Scan an identity document
DocScan
Generate an embedding
text-embed · img-embed
Retrieve tool definitions
llm-tools
Execute a supported tool call
llm-tool-call
Detect and translate visible image text, then render the translation back into the image layout.
Convert supported documents into LLM-ready Markdown, structured JSON, or plain text for downstream application workflows.
A dedicated identity-document scanning product that extracts structured data from supported passports, national IDs, driver's licenses, visas and other identity documents.
Choose by document intent. Use llm-parse when preparing document content for LLM workflows. Use DocScan when the document itself is an identity credential and you need document-specific structured fields.
Generate a numerical embedding vector from text for downstream indexing, similarity and retrieval workflows.
Open the Text Embeddings reference →Generate a numerical embedding vector from an image for downstream indexing and similarity workflows.
Open the Image Embeddings reference →Embeddings are building blocks, not a complete vector database or RAG system. Your application can use returned vectors for semantic search, similarity, retrieval, clustering, classification or indexing and decide how retrieved context is stored and used. See the LLM Tools and Data Parse guide for the deeper retrieval workflow.
Return PixLab-supported tool definitions and schemas in the function-calling format documented by the endpoint.
Submit one or more supported structured tool calls for execution. Tool names must match those defined by the LLM Tools API.
Tool schemas do not imply unrestricted autonomous execution. Your application remains responsible for which tools are exposed, whether a call is authorized, input validation, result validation, and any additional approval or policy controls.
Looking for background removal, media processing, PDF generation, face APIs, or another PixLab endpoint?
Those APIs intentionally remain outside this focused Vision, LLM, document, embedding, and tool reference.
PixLab endpoints do not all use the same HTTP method, parameters, request body, or response shape. The safest integration path is to choose the capability first and treat its dedicated documentation as the implementation source.
Start with the task chooser or compare neighboring APIs in the catalog.
Confirm the documented request method, parameters, accepted inputs, and response fields.
Create or retrieve your API key from the PixLab Console before integrating the endpoint.
Implement the endpoint exactly as its reference specifies rather than applying a universal request template.
Handle the endpoint's documented success and error states before using returned data downstream.
Add your own validation, authorization, retries, error handling, and workflow controls where appropriate.
These products are related but not interchangeable. This page covers the APIs developers compose into their own systems. The products below serve different use cases.
Windows desktop AI workspace for writing, screenshots, OCR, documents, recordings, reusable skills, connected tools, and reviewable multi-step workflows.
Developer building blocks spanning vision, documents, language models, media processing, embeddings, tools, and more.
Compare the full API catalog →Browser-based document and vision tools for workflows that do not require building an API integration first.
Open the Workspace guide →Dedicated identity-document scanning and structured extraction for supported passports, IDs, licenses, visas, and related documents.
Open the DocScan overview →Short answers for choosing an endpoint. Implementation details stay in the endpoint documentation. Can't find the answer you need? Contact support.
Use the query endpoint. It accepts an image together with a natural-language question or instruction and returns an answer based on the submitted visual content.
Use the OCR endpoint for optical character recognition and text extraction from images. For identity documents that require document-specific structured extraction, use DocScan instead.
Use llm-parse. It converts supported documents into LLM-ready Markdown, JSON, or text. The dedicated LLM Tools and Data Parse guide covers the broader parsing and retrieval workflow.
text-embed generates a numerical vector from text, while img-embed generates a numerical vector from an image. Your application can use those vectors for workflows such as similarity, retrieval, semantic search, clustering, classification, or indexing.
llm-tools returns PixLab-supported tool definitions and schemas; retrieving those definitions does not execute a tool. llm-tool-call accepts supported structured tool calls for execution. Tool names and schemas must match PixLab-supported definitions, while application-level authorization and validation remain the developer's responsibility.
Open the dedicated documentation linked from each endpoint in this catalog. Those endpoint pages are the source of truth for supported methods, parameters, request formats, response fields, and available code examples.
Open the PixLab Console ↗ to create or retrieve your API key, then follow the authentication instructions in the documentation for the endpoint you are integrating.
Think-Act is a related Windows desktop AI workspace for writing, screenshots, OCR, documents, recordings, reusable skills, connected tools, and reviewable multi-step workflows. PixLab APIs are developer building blocks for applications; this page does not imply that every API is automatically integrated with Think-Act.
Generate an API key, open the endpoint documentation for your task, and make your first request.