screencastgen

Audio Pipeline

Extract → chunk → synthesize → concatenate into a single audio file.

Source: screencastgen/pipelines/audio.py

---

Function

run_audio_pipeline(request, reporter, backend_factory) -> PipelineRunResult

Parameters:

Param	Type	Description
request	AudioPipelineRequest (Pipeline Types)	Pipeline configuration
reporter	PipelineReporter (Pipeline Events)	Progress output
backend_factory	callable (optional)	Override for backend creation

Returns: PipelineRunResult with exit_code, output_name, output_path

---

Steps

1. Create TTS backend       ← TTS Registry
2. Extract and chunk         ← Pipeline Common → Extractor + Text Processing
3. Validate chunks           ← Pipeline Common
4. Synthesize audio chunks   ← Pipeline Common → TTSBackend.synthesize()
5. Concatenate               ← Concatenator (unless --no-concat)

---

Configuration

Key fields from AudioPipelineRequest:

Field	Type	Default	Description
pdf	str	—	Input document path
backend	str	"qwen"	TTS backend name
device	str	"auto"	Compute device
no_concat	bool	False	Skip concatenation
output_dir	str	"audio"	Output directory
clean	bool	False	Clear previous state

---

Dependencies

Audio Pipeline
├── Pipeline Common   (extract, validate, synthesize)
├── Pipeline Events   (PipelineReporter)
├── Pipeline Types    (AudioPipelineRequest, PipelineRunResult)
├── Concatenator      (merge audio files)
└── TTS Registry      (create backend)

---

See Also

• Pipeline Overview — Shared pipeline design
• Highlight Pipeline — Extends audio with alignment + video
• Data Flow — Audio pipeline flow diagram

This site is open source. Improve this page.