> ## Documentation Index
> Fetch the complete documentation index at: https://docs.freestylevoice.com/llms.txt
> Use this file to discover all available pages before exploring further.

# Formats

> Reshape your transcript based on the app you're dictating into.

The way you want text to look depends on *where* it's going. A message to Slack should stay casual; an email in Gmail wants a greeting and a sign-off; a commit message wants a terse imperative line. **Formats** apply the right style automatically based on the app or site you're dictating into.

Dictate the same rambled thought into two apps and get two appropriately-shaped results — without changing a setting.

## How it benefits your output

A format attaches **instructions** to an **app pattern**. When the active app matches, those instructions steer the [AI cleanup](/features/cleanup) step, so the model doesn't just tidy your words — it shapes them for the destination.

For example, an "Email" format matching `mail.google.com` might instruct: *"Format as a professional email with a greeting, body, and sign-off."* Say "hey can you send me the report thanks" into Gmail and you get a proper email; say it into Slack and it stays a quick message.

## Add a format

<Steps>
  <Step title="Open Formats">
    Find it in the sidebar and click **Add format**.
  </Step>

  <Step title="Label it">
    Give it a name like "Email" or "Commit message".
  </Step>

  <Step title="Set an app pattern">
    Enter a pattern that matches the app name, URL, or window title. Patterns are **pipe-separated** and **case-insensitive**, and match if any piece appears anywhere in the context — for example `mail.google.com | outlook | Spark`.
  </Step>

  <Step title="Write the instructions">
    Describe how the text should be shaped, like "Format as a professional email with a greeting, body, and sign-off."
  </Step>
</Steps>

## Matching rules

* **Substring match.** A pattern piece matches when it appears anywhere in the app context (name, URL, or window title). `mail.google.com` matches the full Gmail URL.
* **Case-insensitive.** Casing never matters.
* **Your rules win.** Custom formats take priority over the built-in defaults, so you can override a default's behavior for a given app.
* **First match applies.** Only the first matching format is used per dictation.

## Defaults and custom rules

Freestyle ships with sensible default formats you can use as-is or tweak. Editing a default turns it into a custom rule (it gets a **CUSTOM** badge). **Reset to defaults** removes all your custom rules and restores the built-in set.

<Note>
  Formats work by steering the cleanup model, so they only take effect when [AI cleanup](/features/cleanup) is enabled. With cleanup off, your transcript isn't reshaped and formats have nothing to hook into.
</Note>

## Where it fits

Formats influence the [cleanup](/features/cleanup) stage — after transcription, before the [dictionary](/features/dictionary) applies its exact replacements. They decide *how* the text is written; the dictionary then swaps in any exact phrases you've defined.