Skip to content

Podcast Modes

Jump to bottom
Muhammad Naufal Rizqullah edited this page Jun 26, 2026 · 1 revision

🎙️ Podcast Modes

OpenSource Clipping provides two intelligent rendering modes specifically designed for podcast-style videos with multiple speakers. Both modes support 3+ speakers across multiple scenes.

Overview

Feature --split-screen --camera-switch
Visual Layout Split (Top-Bottom) Full Screen (Switching)
Dynamic Mode ✅ --dynamic-split (Auto-toggle) ✅ Always Dynamic
Trigger Source Audio or Visual Audio Only (Diarization)
Reaction Shots ✅ Both speakers visible ❌ Only active speaker
Requirement Optional HF_TOKEN HF_TOKEN (Required)

Mode 1: Split-Screen (--split-screen)

Divides the screen into panels to show multiple speakers simultaneously.

┌──────────────┐
│  Speaker A   │  ← Top panel (face-tracked)
│   (Active)   │
├──────────────┤
│  Speaker B   │  ← Bottom panel (dimmed if inactive)
│  (Inactive)  │
└──────────────┘

Basic Usage

# Permanent split-screen (audio-based, needs HF_TOKEN)
python main.py --url "PODCAST_URL" --split-screen

# Dynamic split (auto-toggle between full and split)
python main.py --url "PODCAST_URL" --split-screen --dynamic-split

# Visual-based trigger (NO TOKEN REQUIRED!)
python main.py --url "PODCAST_URL" --split-screen --dynamic-split --split-trigger face

Split Trigger Modes

Trigger Flag Token Required Dimming Effect How It Works
Diarization (default) --split-trigger diarization ✅ HF_TOKEN ✅ Yes Uses audio to detect who is speaking
Face --split-trigger face ❌ No ❌ No Uses visual face count (2+ faces = split)

Tip: Use --split-screen --dynamic-split --split-trigger face for the fastest rendering without any API tokens.

Split-Screen Optimization

Auto-Zoom (--split-auto-zoom)

Dynamically adjusts the zoom level of each panel to keep framing tight on the speaker and exclude other faces. Essential when speakers sit close together.

python main.py --url "PODCAST_URL" \
  --split-screen --dynamic-split --split-trigger face \
  --split-auto-zoom --split-v-align 0.4
Parameter Flag Default Description
Manual Zoom --split-zoom 1.0 Static zoom factor for all panels
Vertical Align --split-v-align 0.5 0.0 = top, 0.5 = center, 1.0 = bottom
Auto-Zoom --split-auto-zoom off AI-driven zoom per panel
Max Zoom --split-max-zoom 2.5 Maximum zoom limit

Mode 2: Camera Switch (--camera-switch)

Mimics professional editing by focusing only on the active speaker in full screen, switching between them cinematically.

Scene 1: Speaker A talking         Scene 2: Speaker B responds
┌──────────────┐                   ┌──────────────┐
│              │                   │              │
│  Speaker A   │   ──── CUT ────→  │  Speaker B   │
│  (Full 9:16) │                   │  (Full 9:16) │
│              │                   │              │
└──────────────┘                   └──────────────┘

Basic Usage

# Standard camera switch
python main.py --url "PODCAST_URL" --camera-switch

# With longer hold duration (prevents flickering)
python main.py --url "PODCAST_URL" --camera-switch --switch-hold-duration 3.0

# 3-speaker podcast
python main.py --url "PODCAST_URL" --camera-switch --diarization-speakers 3

How Camera Switch Renders

Scenario Rendering
1 speaker active Full 9:16 crop following that speaker's face
2+ speakers active, same scene Blurred Pillarbox — original 16:9 frame centered with blurred background
2+ speakers active, different scenes Stays focused on current speaker (no pillarbox)
No one speaking Holds on last active speaker

Hold Duration

The --switch-hold-duration parameter (default: 2.0 seconds) prevents the camera from switching too quickly during rapid dialogue. Increase it for calmer conversations, decrease it for high-energy debates.

Multi-Speaker Support (3+ Speakers)

Both modes support 3 or more speakers across multiple scenes.

Speaker Detection

Mode Flag Description
Auto-detect --diarization-speakers auto (default) Visual scan of 20 frames to auto-count speakers
Fixed count --diarization-speakers 3 Manually specify exact speaker count 
# 3-speaker podcast with camera switch
python main.py --url "PODCAST_URL" \
  --camera-switch \
  --diarization-speakers 3

How 3+ Speakers Work in Split-Screen

  • The screen is always divided into 2 panels (top & bottom)
  • Panels swap between speakers based on who is active
  • Each speaker has their own frozen frame fallback — when not visible in the current scene, the panel shows their last valid frame

Requirements

HuggingFace Token Setup

  1. Create an account at HuggingFace
  2. Create an access token at Settings → Tokens
  3. Accept the Pyannote Speaker Diarization 3.1 agreement
  4. Add to your .env file: 
    HF_TOKEN=hf_your-token-here

Note: --split-trigger face mode does not require HF_TOKEN.

Recommended Configurations

Quick & Token-Free

python main.py --url "PODCAST_URL" \
  --split-screen --dynamic-split --split-trigger face \
  --face-detector yolo

Best Quality (With Token)

python main.py --url "PODCAST_URL" \
  --split-screen --dynamic-split \
  --split-trigger diarization \
  --split-auto-zoom \
  --face-detector yolo

Cinematic Interview

python main.py --url "PODCAST_URL" \
  --camera-switch \
  --switch-hold-duration 2.5 \
  --face-detector yolo

See Also

🎬 OpenSource Clipping

Features

Advanced

Help

Clone this wiki locally