G: Nice project—definitely doable, but there are a few major time sinks and technical annoyances depending on how faithful you want to be. Here’s how I’d approach it, broken down by phase, with notes on shortcuts and complications:

🧩 Phase 1: Source Acquisition

1. Original Films (Unriffed)

• Goal: Find the highest-quality version (preferably the original aspect ratio, uncut).
• Sources:
  • Archive.org (especially for public domain films)
  • Boutique DVD releases (e.g., Shout! Factory, Kino, Mill Creek)
  • Rare film collector forums or torrent trackers (for non-commercially available versions)

2. MST3K Episodes

• Goal: Get digital rips with the best audio fidelity, even if video is VHS-tier.
• Sources:
  • Official DVD sets
  • Archive.org or trading communities (many fans have ripped every episode)

🎬 Phase 2: Timeline Syncing

3. Extract Riff Audio

• Use FFmpeg to demux the MST3K episode:

  ffmpeg -i mst3k_episode.mkv -map 0:a -vn mst3k_audio.wav

4. Riff Track Isolation

• You may want to isolate the riffing from the movie audio:
  • MST3K audio is usually a composite mix of movie + riffers.
  • Options:
    • Accept the bleed (easiest).
    • Phase inversion: if you have the clean movie audio from the original film, invert and subtract (can cancel the film audio).
    • AI vocal isolation (e.g. Adobe Podcast Enhance, Demucs) to split the voices and background.

⏱ Phase 3: Audio/Video Sync

5. Aligning Audio to Original Video

• This is the most labor-intensive part due to:
  • Scene cuts, trims, or reordering in MST3K versions.
  • Intros/outros and host segments breaking up the film.
• Workflow:
  • Open both videos side by side in DaVinci Resolve, Premiere, or even Kdenlive.
  • Identify scene anchors (use a slate or visual marker at key riffed moments).
  • Use slip edits or nested sequences to preserve sync between riff and visual reference.

🔧 Shortcut Tip: Align first 5 riffable moments, and interpolate from there. Most MST3K edits are block trims—not micro-edits.

🗂 Phase 4: Exporting & Track Management

6. Export with Multiple Audio Tracks

• Format: MKV or MP4 with:
  • Track 1: Original audio
  • Track 2: MST3K riff audio

ffmpeg -i original_video.mp4 -i mst3k_audio.wav \
-map 0:v -map 0:a -map 1:a \
-c:v copy -c:a aac -shortest \
metadata:s:a:1 language=eng \
output_with_tracks.mkv

7. Optional Subtitles

• If you isolate riff dialog, you could create subtitles for just the riffs.
• Tools: Aegisub, Whisper (for auto-transcribe), Subtitle Edit.

⚠ Complications and Gotchas

8. Timing Mismatches

• MST3K edits sometimes jump over “dead air” or add pauses. Riffers may react to timing beats that don’t exist in the original version.

9. Different Cuts

• Some MST3K movies use different prints than available now (TV versions, VHS masters). Be prepared to hunt matching frames or rebuild scenes.

10. Audio Licensing Weirdness

• For personal archiving, you’re fine. But the riffs are copyrighted, so public remixing would trigger IP conflicts.

🧠 Alternate Approaches / Time-Savers

A. Use Rifftrax Style Instead

• Overlay the riff audio as a downloadable, time-coded track (like Rifftrax does), requiring the user to provide the movie.
• Pros: Zero video editing.
• Cons: User sync required.

B. Automated Scene Matching

• Use tools like SceneCut or scenedetect to segment both versions and match cuts programmatically.

Let me know how much you want to automate vs. finesse, and I can help design a semi-automated workflow—especially if you're doing this for multiple episodes.

G: Absolutely—let’s break it down into two parts: scene matching software and audio editing tools on Fedora.

🎞️ PART 1: Scene Matching Software

🔍 Goal:

Detect where scenes begin and end in both the original film and the MST3K version so you can:

• Compare timestamps
• See what was cut or trimmed
• Align segments without manual frame-scrubbing

🛠️ Recommended Tools

1. PySceneDetect

• Homepage: https://pyscenedetect.readthedocs.io/
• What it does: Automatically detects scene boundaries using content or threshold-based detection.
• How it works:

  scenedetect -i your_video.mp4 detect-content list-scenes

• Pros:
  • CLI-based and scriptable
  • Outputs a list of scene timestamps
  • Works with ffmpeg to extract clips
• Cons:
  • Can’t match scenes between videos natively—just identifies scenes per video

2. SceneCut

• GitHub: https://github.com/Breakthrough/SceneCut
• What it does: Extends scene detection with machine learning
• Status: Less actively maintained, but offers deep-cut detection
• Note: This is more experimental. PySceneDetect is more practical for your workflow.

3. Match Cuts Manually With OCR or Hashing (optional)

• Use frame hashing (via ffmpeg + perceptual hash) or OCR-based fingerprinting to match scene start/end frames across two different video versions. This is complex but doable with scripting and tools like ImageHash or OpenCV.

🧠 Suggested Workflow

1. Run PySceneDetect on both videos.
2. Export CSV with timestamps.
3. Create a spreadsheet or script to find overlapping/matching timestamps.
4. Use matched scene blocks to align riff audio to original film.

If needed, I can help write a script to do overlap testing and rough scene pairing.

🔊 PART 2: Audio Software for Fedora

You’ll want a tool that allows:

• Multi-track waveform viewing
• Easy region selection
• Phase inversion
• Cut, split, and export per segment
• (Optional) Spectrogram view for deep cleanup

🎚️ Best Options

1. Audacity

• Install: sudo dnf install audacity
• Pros:
  • Intuitive waveform display
  • Supports multi-track editing
  • You can invert one track, align it with another, and mix to subtract
  • Built-in effects: Invert, Normalize, Noise Removal, etc.
• How to do phase inversion:
  • Load original movie audio and MST3K track
  • Duplicate original movie audio
  • On one copy: Effect → Invert
  • Mix the inverted with the MST3K audio to attempt removal
• Cons: Not non-destructive (changes are committed unless you manually undo)

2. Ardour

• Install: sudo dnf install ardour
• Pros:
  • Professional-grade DAW (digital audio workstation)
  • Fully non-destructive, with advanced routing and automation
  • Great for syncing tracks over long-form content
• Cons:
  • Steeper learning curve
  • Requires JACK or PipeWire audio config to be set up properly

3. Reaper (via WINE or native Linux tarball)

• Not open source, but extremely efficient and pro-tier.
• Free evaluation forever (uncrippled).
• Excellent waveform visualization and phase control.
• Available as native Linux binary: https://www.reaper.fm/download.php

⚙️ Short Summary of Phase-Inversion Workflow in Audacity

1. Load both original and MST3K audio.
2. Select original movie audio → Effect > Invert.
3. Play both tracks together; ideally the movie audio will cancel out.
4. What’s left should mostly be riffs + some artifacts.
5. Export that as a new clean riff track (or use for syncing).

⚠️ Caveats for Inversion

• It only works well if the MST3K used the exact audio print you're subtracting.
• If the riff track was mixed with compression or timing differences, results will be partial.
• You may need manual EQ or noise gating to remove bleed that wasn't cancelled cleanly.

Let me know how deep you want to go into automating the alignment or scripting scene pair matching. I can help design a toolchain that lets you rip through batches if you're doing this at scale.

G: Yes—Cool Edit Pro 2000 can usually be run just fine on modern Linux via WINE, and given your experience, that might be the most intuitive tool for your workflow.

Here’s a detailed step-by-step guide tailored for Fedora to get Cool Edit Pro 2000 running smoothly under WINE, with notes for stability and audio compatibility.

🛠️ STEP 1: Install WINE (if not installed)

sudo dnf install wine

Optionally install winetricks for extra configuration help: