Job Description

Transcribe and annotate long-form podcast recordings to support the training and evaluation of speech recognition models. You’ll combine careful listening with consistent speaker labels, timestamps, and audio-quality checks.

## About this project

Freya focuses on accurately segmenting and transcribing speech in podcast recordings that typically run for 30–60 minutes or longer. The resulting data helps train and evaluate voice recognition and speech-to-text models across many languages and locations.

## What you’ll do

* Transcribe long-form podcast recordings according to project guidelines.
* Identify speakers and apply consistent speaker labels.
* Segment speech and add accurate timestamps.
* Record the perceived gender of speakers when required by the guidelines.
* Flag unclear or distorted speech, strong accents, incorrect languages, and synthetic or generated audio.
* Review and correct pre-annotated speech segments when required.

## Requirements

* Native-level proficiency in Japanese (Japan).
* Strong listening skills and close attention to detail.
* Previous transcription or speech annotation experience.
* A computer with a stable internet connection.

## Project details

* **Location:** Remote within the eligible language and location options.
* **Task length:** Podcast recordings generally run for 30–60 minutes or longer.
* **Availability:** Tasks are released weekly on a first-come, first-served basis. Volume may vary by location and batch.

## Compensation

$12 USD per hour

## Register here:

##

Job Tags

Similar Jobs