Biblicus

Make your documents usable by your assistant, then decide later how you will search and retrieve them.

Speech-to-Text (STT)

Audio transcription extractors for converting spoken content into text.

Speech-to-Text Extractors

Overview

Speech-to-text extractors transcribe audio from video and audio files. They are ideal for:

The raw audio bytes remain unchanged in the corpus; only transcribed text is stored in extraction results.

Available Extractors

stt-openai

OpenAI Whisper API for audio transcription:

Installation: pip install biblicus[openai]

Best for: General transcription, multi-language content, OpenAI ecosystem integration

stt-deepgram

Deepgram Nova-3 for fast, accurate transcription:

Installation: pip install biblicus[deepgram]

Best for: High-accuracy transcription, speaker diarization, professional content

stt-aldea

Aldea Speech-to-Text API for audio transcription:

Installation: pip install biblicus[aldea]

Best for: Aldea-hosted transcription, Deepgram-compatible workflows

deepgram-transform

Render Deepgram structured metadata into text:

Best for: Diarized filtering, channel selection, and structured transcript rendering

Choosing an Extractor

Use Case Recommended Notes
General transcription stt-deepgram Better accuracy, formatting
Multi-language content stt-openai More languages supported
Speaker identification stt-deepgram Has diarization feature
Translation to English stt-openai Built-in translation
Cost-sensitive stt-deepgram Competitive pricing
OpenAI workflow stt-openai Single API key
Aldea / Deepgram-shaped API stt-aldea Aldea-hosted, same response shape as Deepgram

Performance Comparison

OpenAI Whisper

Deepgram Nova-3

Common Patterns

Fallback Chain

Try Deepgram first, fall back to OpenAI:

extractor_id: select-text
config:
  extractors:
    - stt-deepgram
    - stt-openai

Language-Specific Routing

Route by media type or use overrides:

extractor_id: select-smart-override
config:
  default_extractor: stt-deepgram
  overrides:
    - media_type_pattern: "audio/.*"
      extractor: stt-deepgram
    - media_type_pattern: "video/.*"
      extractor: stt-openai

Speaker Diarization

Use Deepgram with diarization enabled:

extractor_id: stt-deepgram
config:
  diarize: true
  smart_format: true

Authentication

Both extractors require API keys:

Environment Variables

export OPENAI_API_KEY="your-openai-key"
export DEEPGRAM_API_KEY="your-deepgram-key"

Configuration File

Add to ~/.biblicus/config.yml:

openai:
  api_key: YOUR_OPENAI_KEY

deepgram:
  api_key: YOUR_DEEPGRAM_KEY

aldea:
  api_key: org_YOUR_ALDEA_KEY

Supported Audio Formats

Both extractors support common audio formats:

See Also