Skip to content

edit — Full editing pipeline

The main command. Runs transcription → content detection → plan → render in one shot.

Usage

praisonai-editor edit INPUT [OPTIONS]

All Options

Option Short Default Description
INPUT Input media file (audio or video)
--output -o {stem}_edited.{ext} Output file path
--preset -p podcast Edit preset (see Presets)
--prompt Natural language instruction (AI agent)
--detector auto Content detector: auto · ensemble · ina · librosa · ffmpeg
--demix off Enable Demucs vocal stem separation
--primary-zone off Crop to primary singing zone only
--no-fillers off Keep filler words (don't remove)
--no-repetitions off Keep repetitions
--no-silence off Keep silences
--local off Use offline faster-whisper
--language auto Language code (en, ta, es, …)
--reencode off Re-encode instead of stream copy (slower, better quality)
--no-artifacts off Don't save artifacts to ~/.praisonai/editor/
--verbose -v off Print step-by-step progress

Pipeline overview

flowchart LR
    Input --> Probe
    Probe --> Cache{Transcript\ncached?}
    Cache -->|Yes| Plan
    Cache -->|No| Whisper --> Plan
    Plan --> Detect{Content\npreset?}
    Detect -->|songs_only\nspeech_only\nno_silence| ContentDetect
    Detect -->|podcast\nmeeting\ncourse\nclean| Heuristic
    ContentDetect -->|--demix| Demucs
    ContentDetect --> Render
    Demucs --> Render
    Heuristic --> Render
    Render --> Output

Examples

praisonai-editor edit podcast.mp3 -v
Removes: fillers · repetitions · silences > 1.5s

praisonai-editor edit concert.mp3 \
  --preset songs_only \
  --detector ensemble \
  --demix \
  --primary-zone \
  -v

praisonai-editor edit radio.mp3 --preset speech_only --detector ensemble

praisonai-editor edit meeting.mp4 --preset meeting -v

praisonai-editor edit interview.mp3 \
  --prompt "Remove the intro and any off-topic discussion"

praisonai-editor edit audio.mp3 --language ta -v

pip install "praisonai-editor[local]"
praisonai-editor edit audio.mp3 --local -v

Artifacts saved

~/.praisonai/editor/{filename}/
  ├── probe.json
  ├── transcript.json   ← cached — reused on second run
  ├── transcript.srt
  ├── transcript.txt
  ├── plan.json
  └── content_blocks.json  ← only with --detector

Use --no-artifacts to skip saving.

Python API

from praisonai_editor.pipeline import edit_media

result = edit_media(
    "podcast.mp3",
    output_path="podcast_clean.mp3",
    preset="podcast",
    verbose=True,
)

print(result.success)          # True
print(result.output_path)      # "podcast_clean.mp3"
print(result.plan.removed_duration)  # seconds removed