Random JSON Payload Generator

Mock data is only useful when it looks like the real thing. A field called email holding string1 tells you nothing about how your layout handles a long address. The Random JSON Generator reads a template you paste, infers what each field is from its name and its current value, and refills it with plausible data of the same shape.

Inference is the whole trick. A value matching a UUID pattern is replaced by another UUID, an ISO timestamp by another timestamp, a hex colour by another hex colour, and a semantic version by another version. Field names do the rest, so city, jobTitle, currency and locale each draw from their own vocabulary.

Structure is preserved exactly. Arrays stay arrays of the same length, nesting depth is unchanged, and null stays null, so the output drops into a fixture file wherever the original template came from. Output is syntax-highlighted with key, value, string and number coloured separately, and nothing is uploaded.

Getting Fresh Fixtures From a Template

1. Load one of the built-in templates or paste your own. The presets cover an API envelope with a paginated result set and an IoT sensor reading, which between them exercise nesting, arrays and mixed types.
2. Press Generate. Nothing happens on paste alone, so a large template can be edited without regenerating on every keystroke.
3. Read the stat cards for key count, value count, nesting depth and output size in bytes. Depth is the one to watch: it tells you whether the template is what you thought it was.
4. Press Generate again for a completely different set at the identical shape, which is how you build several fixtures from one schema.
5. Take Copy JSON for the result, or copy the template back out if you edited it in place.

How Big a Template Can Get

• Templates are capped at 50,000 characters, and a longer one returns Template is too large — keep it under 50,000 characters. rather than freezing while it walks the tree.
• The walk stops at 20,000 nodes and 24 levels of nesting. Both are guards against a pathological structure rather than limits you are likely to meet with real fixtures.
• Syntax highlighting is skipped above 120,000 characters of output. The JSON is still correct and copyable; only the colouring is dropped.
• Invalid JSON is reported with the parser's own message, so a trailing comma names its own position rather than producing a generic complaint.
• Inference is a heuristic and will occasionally guess wrong. A field named status holding a number gets a number back, not one of the status strings, because the existing value wins over the name.
• Values are drawn from fixed vocabularies rather than generated from a grammar, so a large array will repeat names and cities. That is fine for a fixture and wrong for a uniqueness test.