---
title: "JSON to Python Dataclass, Pydantic & TypedDict | JSON Lark"
description: "Generate Python dataclasses, Pydantic v2 models or TypedDicts from JSON samples, with snake_case names, aliases and Optional fields. In your browser."
url: https://jsonlark.me/json-to-python/
---

Sheet 03 Python classes

# JSON to Python

Paste JSON samples and get dataclasses, Pydantic v2 models or TypedDicts, with snake_case attributes that still map to your original keys.

## How it works

Each sample is read as a value of the root class. Classes are written in dependency order, so nested classes come first and every annotation resolves.

- **dataclasses**: Fields without defaults come first, then optional ones with `= None`. `from_dict()` maps the original keys to the snake_case attributes and builds nested objects, lists and dicts of them.
- **Pydantic v2**: Renamed attributes get `Field(alias="firstName")` and the model sets `populate_by_name=True`, so both spellings validate.
- **TypedDict**: Keys must match the JSON exactly, so names are never changed; keys that are not identifiers use the functional `TypedDict("Root", {...})` form, and optional keys are `NotRequired`.
- **Syntax**: 3.11+ writes `list[str]` and `str | None`; 3.8+ uses `List`, `Optional` and typing_extensions.

## Questions

**What happens to keys like class or first-name?**

Keywords get a trailing underscore (class_), other names are converted to snake_case (first_name), and the original key is kept in an alias or in from_dict(). Names that would shadow str, list or self get an underscore too.

**Does Pydantic parse nested objects for me?**

Yes. Root.model_validate(data) builds the nested models from plain dicts, using the aliases. With dataclasses, Root.from_dict(json.loads(text)) does the same job.

**Is my JSON uploaded?**

No. The code is generated by JavaScript running in this tab, and nothing you paste leaves it. The site counts page views with Cloudflare Web Analytics, which sets no cookies and never receives what you paste.