Portuguese (Brazil) Audio Transcription Reviewer

About OpenTrain

OpenTrain is the #1 platform for finding and building careers in AI training and data labeling. We help people discover projects, build a unified portfolio, and grow freelance careers teaching AI — all from a single, specialized platform.

About AI training and this work

AI training (also called data labeling or human feedback) is the human side of building modern AI. Contributors perform critical tasks like transcribing audio, evaluating model outputs, and verifying speaker attributions — work that directly shapes how models behave.

This role focuses on speech and transcription QA: you will evaluate machine-generated transcripts and diarization outputs to make sure spoken content, timing, and speaker labels are accurate and faithfully preserved.

The role

You will review AI-generated transcriptions and diarization results for multi-channel Portuguese (Brazil) recordings. The work centers on accuracy targets for transcription, speaker attribution, timestamps, segmentation, and alignment, and requires reading JSON-formatted outputs to confirm tagging and diarization logic.

• Employment type: Contractor, Part-time
• Time requirement: 20+ hours/week
• Experience level: Entry level
• Data type: Audio; label types: Transcription and Evaluation/Rating
• Work location: Remote (worldwide)

What you'll do

• Review AI-generated transcripts and diarization outputs against strict accuracy targets to ensure low error rates.
• Validate turn-level and word-level timestamps for alignment and temporal precision.
• Confirm speaker labels and diarization alignment across multi-channel recordings.
• Ensure transcripts preserve spontaneous, unnormalized speech, including overlaps, interruptions, false starts, and disfluencies.
• Read and interpret JSON-formatted outputs to verify tagging logic, metadata, and diarization markers.
• Identify and flag transcription, segmentation, alignment, and speaker attribution errors for remediation.

Requirements

Candidates must bring strong Portuguese (Brazil) listening and reading skills plus demonstrated ability to evaluate transcription and diarization quality.

• Fluent or native Portuguese (Brazil) listening and reading ability.
• Experience reviewing ASR or AI-generated transcription outputs.
• Ability to judge diarization accuracy and speaker attribution in multi-channel audio.
• Comfort validating word-level and turn-level timestamps and temporal alignment.
• Strong audio-fidelity judgment, attention to detail, and comfort with overlapping dialogue.
• Experience in speech data QA, transcription review, or related audio evaluation work is highly relevant.

Helpful background (nice to have)

• Training or coursework in linguistics, phonetics, or speech-language fields.
• Professional transcription experience or prior work in verbatim/un-normalized transcription.
• Experience evaluating conversational dynamics, disfluencies, and overlap in spontaneous speech.

How it works / Next steps

Apply through your OpenTrain account to be considered. If selected, you'll receive project-specific guidelines, examples, and a QA rubric to follow. Work is done remotely on a contract basis and measured against accuracy targets; you will review JSON outputs and annotated audio as part of standard workflow.

• You will be provided with instructions, examples, and acceptance criteria for each task.
• Work is evaluated for accuracy; feedback loops and calibration tasks may be part of onboarding.
• No specific hourly rate is listed here; project compensation and milestones are shared when you apply.