Pre-recorded Audio

Feature Overview

Complete parameter reference for the batch transcription endpoint. Only model is required.


Core parameters

ParameterTypeDefaultDescription
modelstringrequiredModel name. Use zero-indic.
filefileone ofAudio/video file upload.
urlstringone ofPublic audio/video URL.
language_codestringautoISO 639-1 code (hi, ta, en) or full language name.
output_scriptstringautoTransliterate output. Values: auto, Devanagari, Tamil, Telugu, Kannada, Bengali, Latin, ITRANS.
word_timestampsboolfalseAdd per-word start/end times and alignment score to every segment.
projectstringnoneSpeaker library namespace. Also tags requests for analytics.

Speaker intelligence parameters

ParameterTypeDefaultDescription
enable_diarizationboolfalseLabel each segment SPEAKER_00, SPEAKER_01, etc.
enable_speaker_identificationboolfalseMap SPEAKER_XX labels to registered names. Requires enable_diarization=true.
enable_emotion_diarizationboolfalseAdd emotion to each segment. Requires enable_diarization=true.
enable_sentiment_analysisboolfalseSentiment analysis using Gemini. Result in nlp_analysis.sentiment.
enable_intent_detectionboolfalseIntent classification using Gemini. Result in nlp_analysis.intent.
intent_choicesJSON arraynoneConstrain intent to specific labels.

Transcript enrichment parameters

ParameterTypeDefaultDescription
enable_summarizationboolfalseGenerate transcript summary using Gemini. Result in nlp_analysis.summary.
summary_max_lengthint150Approximate maximum word count of the summary.
output_languagestringnoneTranslate full transcript using Gemini. Result in nlp_analysis.translation.
enable_keyterm_normalizationboolfalseNormalise domain-specific terms using Gemini. Result in nlp_analysis.normalized_text.
keyterm_keywordsJSON arraynoneFocus normalisation on specific terms.

Formatting & redaction parameters

ParameterTypeDefaultDescription
enable_profanity_hashingboolfalseDetect and mask profanity using Gemini. Replaces with **** in text and segments.
hash_keywordsJSON arraynoneMask specific words/phrases using regex (no LLM). Replaces with ****.