[Hiring] Welsh Transcriber @Veroa.ai

Role Description

We're looking for an experienced Welsh Transcriber to support a conversational audio transcription project developing voice datasets for AI labs. You'll review and correct automatic speech recognition (ASR) transcripts of Welsh-language audio, ensuring the final output meets a word error rate (WER) of 5% or less. This is a verbatim transcription project, not standard subtitling. Accuracy, consistency, and attention to linguistic detail are essential.

• Correct automatic speech recognition (ASR) drafts of Welsh conversational audio.
• Ensure final transcripts meet a word error rate of 5% or less.
• Listen to Welsh conversational audio and correct the ASR draft verbatim.
• Keep fillers, repetitions, stuttering, false starts, incomplete sentences, dialect, and backchannels.
• Do not normalize grammar.
• Keep English speech in English; do not translate it into Welsh.
• Correct ASR mistranslations, including segments where ASR converted English into Welsh.
• Label speakers and correct start and end timestamps to 0.01 seconds.
• Split segments at natural pauses, keeping them within 30 seconds where possible.
• Deliver JSON segments with speaker, start, end, and text, without tags like (laughter) or (cough).
• Human-review every file in full; ASR is only a starting point.

Qualifications

• Fluent in Welsh, with excellent listening comprehension and written language skills.
• Previous professional transcription experience.
• Strong attention to detail, particularly with regional accents, dialects, and fast conversational speech.
• Comfortable following detailed transcription and formatting guidelines.
• Able to work independently and consistently meet quality and delivery requirements.
• Available to work an 8-hour weekday schedule during the anticipated project period.
• Comfortable using digital transcription tools and working with structured files.

Requirements

• ASR Correction: Experience reviewing and correcting automatic speech recognition drafts is helpful, but not required.
• Technical Exposure: Familiarity with speaker labeling, timestamp alignment, or JSON tools is helpful, but not required.
• Domain Knowledge: Background in linguistic data, AI training datasets, or professional transcription for speech technology is helpful, but not required.

Benefits

• High visibility, fast-paced learning, and ownership over real client outcomes.
• Embrace modern technology, AI workflows, and smart automation internally and externally.
• Work with a team that values output, creative problem-solving, and work-life harmony.