Crate scribble 

scribble — a small, focused transcription library built on top of Whisper.

Overview

Scribble provides a clean, idiomatic Rust API for audio transcription, designed to work equally well in CLI tools and long-running services.

At a high level, Scribble wires together:

• Media demuxing and audio decoding (via Symphonia)
• Audio normalization and resampling (mono, 16 kHz)
• Optional Voice Activity Detection (VAD)
• Whisper inference
• Pluggable output encoders (JSON, VTT, etc.)

The library emphasizes:

• Explicit control flow
• Streaming-friendly design
• Clear separation of concerns
• Minimal surprises for callers

Most consumers should start with scribble::Scribble.

Modules

audio_pipeline

Audio resampling, downmixing, and chunk emission pipeline. Audio normalization pipeline for Scribble.

ctx

Whisper model loading and context management.

decode

Codec-level decode helpers. Decoder helpers built on top of Symphonia.

decoder

Streaming-friendly audio decoding and normalization helpers. Stream-decode media (audio/video containers) into Whisper-friendly mono f32 @ 16kHz, emitting fixed-size chunks via a callback.

demux

Low-level demux helpers (container probing, packet iteration). Demux helpers for Symphonia.

json_array_encoder

JSON array encoder.

logging

Logging configuration and control.

opts

User-configurable transcription options.

output_type

Output format selection.

scribble

User-facing transcription entry point and orchestration logic. High-level API for running transcriptions with Scribble.

segment_encoder

Shared encoder trait definitions.

segments

Segment data structures and transcription helpers.

vad

Voice Activity Detection (VAD) utilities and policies.

vtt_encoder

WebVTT encoder.

wav

WAV helpers (primarily for tests and simple inputs).