Audio Processing Pipelines

Indexify is capable of ingesting and processing audios. You can build pipelines that perform one or more of these functions -

• Transcription
• Summarization
• Topic Extraction
• Indexing

Each of these examples are organized as -

1. Extraction Graph/Pipeline Description
2. Pipeline Setup Script
3. Upload audio files, and retrieve the artifacts of the pipeline such as transcripts, summary, topics, etc.

Transcription Models

Indexify provides the following transcription extractors out of the box -

• OpenAI Whisper
• Apple's Whisper MLX(MacOS)
• Groq Whisper API
• DistilWhisper x PyAnnotate based speaker diarization

You can list all the ASR extractors by running the command indexify-extractor list

There are a ton of other ASR APIs out there, which we haven't integrated with. We would love the community to contribute or you can write a custom extractor to call into any API.

Indexing

The indexing example uses LanceDB by default, but you can swap it out with any other VectorDB that Indexify supports - Qdrant or PgVector.

We are using the embedding model - minilm-l6 in this example, you can list the other available extractor by running the command - indexify-extractor list