OpenViSTAI · Agents · Blockchain · Crypto
Submit a project
Menu
Audio & Voice AI

Kaldi Offline Transcriber

Kaldi Offline Transcriber is a self-hosted Estonian speech-recognition system that processes local audio recordings into text and structured outputs containing speaker, word, and timing information.

Explore the practical guide ↓

Independent OpenViST project profile, based on linked sources.

Kaldi Offline Transcriber — OpenViST project profile in Audio & Voice AI.
OpenViST project illustration. See the linked sources for product details.
Understand the project

Practical overview

When it fits

Use it when you can run a Linux-based transcription workflow locally and need Estonian speech recognition with structured timing or speaker information.

Who it is for

  • People who need to transcribe long Estonian audio recordings.
  • Teams or individuals processing interviews, meetings, or broadcasts on a Linux server.
  • Users who need speaker labels or word- and sentence-level timing in transcription outputs.

Documented uses

  • Create a text transcript from an Estonian audio file.
  • Separate speech from non-speech and identify speaker turns.
  • Produce transcript data with speakers, words, and timings.
  • Generate subtitle-oriented or time-aligned outputs from recorded speech.
Before you choose

Requirements & limits

The project is a public GitHub repository with Linux and Docker deployment documentation. Its README describes local installation and processing rather than a hosted service.

  • Model initialization needs around 8 GB of RAM; transcription after initialization needs around 2 GB.
  • The README lists WAV, MP3, OGG, MPG, and M4A inputs.
  • Documented outputs include text, TRS, CTM, compound-word CTM, and JSON with speaker, word, and timing information.
  • The workflow includes language identification that filters non-Estonian utterances before decoding.
  • A batch upload and job-management interface is an experimental adaptation idea, not a documented current capability.

Check current requirements in the official documentation. Source-based guidance does not mean OpenViST has run or independently tested this project.

Source-based practical guide

Transcribe a local audio file

Generate a text transcript from an audio recording using the repository’s documented workflow.

Before you start

  • A Linux server with the required components installed as described by the repository’s Dockerfile.
  • Around 8 GB of RAM to initialize the models and around 2 GB for transcription after initialization.
  • An audio file; the README lists WAV, MP3, OGG, MPG, and M4A.

Steps

  1. Place the audio file in the src-audio directory.

  2. Run the command below, replacing `` with the audio filename without its extension.

    Documented command
    make build/output/.txt

Expected outcome

The pipeline generates a text transcript under `build/output`.

Guide limitations

  • The documented command produces a text transcript; this guide does not cover exporting other formats.
  • Processing speed varies by machine; the README gives an approximate 0.6-times-real-time figure for a one-CPU server.
Documentation 1 ↗
Evidence and transparency

Sources & review

Official repository ↗
Source record details
Source availability
Active
Sources checked
15 October 2026
Original website title
alumae/kaldi-offline-transcriber · Official homepage title · retained in its source language
How we review sourcesSuggest a correction