Spanish Podcast Transcription & Annotation -- 2
Budget: ₹2,500 – ₹0 INR
The objective of this project is to produce high-quality edited Spanish podcast transcriptions with precise speaker segmentation and timestamps. Audio files range from 2 to 45 minutes and may include multiple speakers, accents, and dialects.
⚠️ Before starting the full batch, you must complete a mandatory 2-minute test audio to verify style, timestamp accuracy, speaker labeling, and overall quality.
Platform & Access
All work must be completed on a designated transcription platform
Platform link, user ID, and password will be provided after selection
Transcription, annotation, and edits must be done directly inside the assigned platform
Transcription Requirements
Edited transcription (not raw verbatim):
Correct grammar and punctuation
Remove unnecessary filler words only when it improves readability
Preserve full speaker intent and meaning
Spanish only (original register preserved; no English translation)
Segmentation & Timestamps
Insert timestamps at every speaker change
Timestamp format: [hh:mm:ss] (final delivery)
Each segment must include:
Speaker label (e.g., HOST, GUEST, Speaker A/B)
Accurate start/end times (do not cut words)
Avoid overly long segments (ideal: ≤25–30 seconds)
Segments from the same speaker must never overlap
Speaker & Metadata Annotation
Use consistent speaker identifiers throughout the episode
Annotate when confident:
Language
Locale (e.g., es-LATAM)
Accent / dialect (best educated guess if unclear)
Non-Speech, Emotion & Emphasis
Include audible non-speech events:
(risas), (suspiro), (música), etc.
Annotate emotions only when clearly expressed
Mark exaggerated emphasis sparingly
Accuracy & Quality Bar
Minimum 98% accuracy
Zero unintended omissions
Fully proof-read and spell-checked
⚠️ Before starting the full batch, you must complete a mandatory 2-minute test audio to verify style, timestamp accuracy, speaker labeling, and overall quality.
Platform & Access
All work must be completed on a designated transcription platform
Platform link, user ID, and password will be provided after selection
Transcription, annotation, and edits must be done directly inside the assigned platform
Transcription Requirements
Edited transcription (not raw verbatim):
Correct grammar and punctuation
Remove unnecessary filler words only when it improves readability
Preserve full speaker intent and meaning
Spanish only (original register preserved; no English translation)
Segmentation & Timestamps
Insert timestamps at every speaker change
Timestamp format: [hh:mm:ss] (final delivery)
Each segment must include:
Speaker label (e.g., HOST, GUEST, Speaker A/B)
Accurate start/end times (do not cut words)
Avoid overly long segments (ideal: ≤25–30 seconds)
Segments from the same speaker must never overlap
Speaker & Metadata Annotation
Use consistent speaker identifiers throughout the episode
Annotate when confident:
Language
Locale (e.g., es-LATAM)
Accent / dialect (best educated guess if unclear)
Non-Speech, Emotion & Emphasis
Include audible non-speech events:
(risas), (suspiro), (música), etc.
Annotate emotions only when clearly expressed
Mark exaggerated emphasis sparingly
Accuracy & Quality Bar
Minimum 98% accuracy
Zero unintended omissions
Fully proof-read and spell-checked