---
title: "ScribeSTT"
description: "Release notes for each ScribeSTT version, newest first"
canonical_url: "https://docs.ardge.com/en/release/scribestt"
---
# ScribeSTT

> Release notes for each ScribeSTT version, newest first

ScribeSTT releases are listed below, newest first.

<u-changelog-versions>
<u-changelog-version badge="Latest" date="2026-09-04" title="ScribeSTT 1.0.0">
<template v-slot:body="">

ScribeSTT 1.0.0 is the first general release of ScribeSTT. ScribeSTT converts audio and video files into transcripts using on-premises speech recognition models, and outputs SRT, TXT and JSON for use by AI agents.

**Initial setup**

The first administrator account is created when the service is first opened.

**Home**

Summarises processing output and recent activity.

- Totals over the last 7 days, 30 days or all time: files processed, words produced, audio duration and estimated time saved
- The most recent transcripts and tasks, with their status and time

**Transcript**

Transcribes individual audio and video files.

- Files are transcribed asynchronously; the page can be closed while processing continues
- The audio language can be set to Chinese, English, Japanese or Korean, or left to the model to detect
- A glossary can be applied to the run to correct recognised terminology
- Transcripts download as SRT, TXT or JSON

**Tasks**

Transcribes folders in batches.

- A task scans a source folder and writes the output to a target folder
- Scan depth limits how far into the folder tree the scan reaches
- A schedule triggers execution automatically at the specified time
- Each execution is recorded with its per-file results

**Datasources**

Connects the remote storage that tasks read from and write to.

- Remote NAS connections over SMB supply the folders that tasks reference; a task target folder requires write access
- The file browser lists folders and files, and supports uploading and deleting
- A datasource referenced by a task cannot be deleted

**Glossary Library**

Corrects terminology during transcription.

- A glossary holds pairs of a commonly misrecognised term and its correct form, applied while the transcript is produced
- Each glossary is assigned a language: Chinese, English, Japanese or Korean
- Terms import in bulk from CSV
- One glossary applies to a transcription run

**Model Management**

Configures the speech recognition model source.

- A single model source is shared across the system, configured with a base URL, API key and model name

**Users Management**

Controls who can use the service and how.

- Accounts are created by an administrator, with the Admin and Member roles determining access
- Disabling an account terminates its sessions immediately
- Personal access tokens give CLI and AI agent clients long-lived credentials, scoped to the actions they may perform and the datasources they may reach

</template>
</u-changelog-version>
</u-changelog-versions>


## Sitemap

See the full [sitemap](https://docs.ardge.com/sitemap.md) for all pages.