Transcriber

What the work is for

Human transcripts teach speech recognition to turn sound into text, and show where it still goes wrong. Automatic systems still miss accents and dialects, which is where a native ear matters.

The work

• Transcribe each assigned recording.
• Add speaker labels and timestamps where the guidelines ask for them.
• Correct machine drafts where the project works that way.

What to expect

• Each language has its own style guide, covering numbers, punctuation and non-speech sounds.
• Some projects want every word and hesitation kept, others a clean, readable text. The guidelines say which.
• Work comes in batches, and each batch is quality-reviewed before it is accepted.
• The work happens in YPAI's review console or in a project tool such as Label Studio.

What you bring

• A good headset.
• A reliable internet connection.

You need

• Native or near-native command of the language you transcribe.
• Fast, accurate typing.
• 18 years or older.

Good to know

• Recordings are real speech, with people talking over each other and background noise.
• Applying does not guarantee a project. You are invited when one needs your language.