Output formats

TranscriptRenderer turns a Transcript into text. Pick a format at run time:

let text = try TranscriptRenderer.render(transcript, as: .srt)
let url = folder.appendingPathComponent("call.\(TranscriptFormat.srt.fileExtension)")
try text.write(to: url, atomically: true, encoding: .utf8)

TranscriptFormat is CaseIterable and has a displayName for pickers.

MarkdownCopy Anchor Link

let markdown = TranscriptRenderer.markdown(transcript, options: .init(
    title: "Weekly sync",
    metadata: [("date", "2026-09-25"), ("tags", "meeting")],
    includeTimestamps: true,
    preamble: "![[weekly-sync.m4a]]"
))

---
date: "2026-09-25"
tags: "meeting"
---

# Weekly sync

![[weekly-sync.m4a]]

[00:00] **Anna:** Morning, everyone.

[00:04] **Speaker 1:** Hi Anna.

• metadata becomes YAML front matter, in the order given. Values are written as quoted strings. Note apps such as Obsidian read it as properties.
• preamble goes after the title, for example an audio embed.
• Timestamps use mm:ss, or h:mm:ss past an hour.

SubtitlesCopy Anchor Link

Segments are split into cues of at most about 7 seconds or 84 characters. A cue ends at the last comma or full stop when there is one, so lines read naturally. Each cue starts with the speaker name.

1
00:00:00,000 --> 00:00:02,000
Anna: Morning, everyone.

JSONCopy Anchor Link

Transcript, Segment, Speaker and Word are Codable, so the JSON output decodes straight back:

let json = try TranscriptRenderer.render(transcript, as: .json)
let decoded = try JSONDecoder().decode(Transcript.self, from: Data(json.utf8))

Times are seconds as Double.

HelpersCopy Anchor Link

• TranscriptRenderer.clock(_:): mm:ss or h:mm:ss.
• TranscriptRenderer.duration(_:): hh:mm:ss.
• transcript.text: all segments joined, without speakers or times.
• transcript.wordCount.