Skip to content

Already have the transcript? Skip straight to the campaign wiki

A transcript is not notes. It is six hours of text nobody will reread. This turns one into the structure you actually wanted, without re-running transcription you already paid for.

Upload a file or paste the text

Import takes a .txt or .md file, or text pasted straight into the box. From there the full chain runs exactly as it would after a recording: entity extraction into the wiki, relationship detection between those entities, and a session recap.

The output is indistinguishable from a recorded session, because it is the same pipeline. An imported session sits in your campaign next to recorded ones, appears in the timeline, and is searchable and answerable the same way.

  1. Create the session you want the transcript to belong to
  2. Upload the .txt or .md file, or paste the text
  3. Entity extraction, relationships and the recap run automatically
  4. Review anything the extractor was unsure about in the wiki review queue

It spends no transcription hours

Transcription hours pay for turning audio into text. An import skips that step, so it costs none of them — no partial charge, no rounding. Your monthly hours are untouched whether you import one transcript or ten.

That is why import is metered separately rather than being free and unlimited: it produces a full wiki, recap and relationship graph, which costs real money to generate, and the hours quota cannot bound something that consumes no hours.

Where the transcript comes from does not matter

Craig recordings you transcribed yourself, output from the Whisper API, a file from another transcription service, or the text log of a play-by-post game all work. There is no required format beyond being readable text — no timestamps needed, no speaker labels expected (we do not do speaker labelling either way).

If you have a transcription pipeline you already like, this is the cheapest way to use the product: keep your pipeline, and let this build the campaign structure on top of it. Running Whisper yourself and recording with Craig both end where this begins.

Importing the same text twice is safe

Re-importing identical text into the same session returns the original transcript instead of running the AI chain again. So a double-click, a flaky connection or an uncertain retry costs nothing and cannot produce a duplicate session.

What this does not do

It does not take subtitle files. .srt and .vtt are rejected with a specific message rather than imported as garbage — they are timing data wrapped around text, and the timings would end up in your wiki. Convert to plain text first.

There is a size ceiling: 500,000 characters, roughly eight hours of speech, which is about the longest transcript the recording side can produce anyway. The binding constraint is entity extraction, which reads the transcript in one pass.

An imported session has no audio, so there is no transcript-to-audio sync — clicking a line will not jump to that moment, because there is no recording to jump into.

And it inherits your transcript's mistakes. If the tool that produced it misheard a name thirty times, the wiki gets that name thirty times. Recording through DM Scribe primes transcription with your campaign glossary specifically to stop that; an imported transcript never had that help.

Which plans include it

On every plan, including the free Adventurer tier — this is not a paid feature. It is metered per month rather than gated: 2 imports a month on Adventurer, 10 on Dungeon Master, unlimited on Guild Master. Going over returns a clear limit message, not a locked door. See the full plan comparison.

Full pricing

The rest of what it does

All features