Study Video Generation Pipeline 🎧📽️

Generate long-form study / focus videos for YouTube using:

• AI music from Mubert (Creator plan / Render)
• AI visuals from DALL·E (via ChatGPT)
• FFmpeg for stitching audio, creating videos, and adding a subtle Ken Burns effect

The goal is to keep the workflow:

• Cheap (leveraging tools you already pay for)
• Semi-automated (minimal manual steps once things are set up)
• YouTube-friendly (resolutions, codecs, licensing considerations)

---

High-Level Flow

1. Generate audio in Mubert
   • Use Mubert Render (Creator plan) to generate several .wav tracks for a given “vibe” (e.g. Italian café, dark cello, fantasy LOTR-ish).
2. Drop the audio files into Audio_Parts/.
3. Generate one or more background images using DALL·E via ChatGPT and save them into Images/.
4. Run the scripts:
   • Option A (N8N / one call): node Script/run-pipeline.js with JSON input (theme, audioDir, imagePath) → merged audio + video + template title/description in one step.
   • Option B (manual): build_audio.sh → concatenates all audio parts; then build_video.sh → combines with a background image and Ken Burns zoom into Outputs/.

---

Project Structure

Suggested layout:

study_video_generation/
  Audio_Parts/      # Raw Mubert audio parts (WAVs)
    part1.wav
    part2.wav
    ...
  Images/           # DALL·E images used as backgrounds
    bg.png
    ...
  Outputs/          # Generated audio + final videos
    audio_merged.m4a
    final_video.mp4
  Script/           # Helper scripts
    build_audio.sh
    build_video.sh
    run-pipeline.js # N8N / single-call pipeline (Node)
    presets.js      # FFmpeg presets (study, epic)
  package.json      # Node entry for run-pipeline.js
  README.md

---

N8N usage (single script)

For automation (e.g. N8N), use the Node pipeline so one call does everything with minimal inputs.

Minimal input (JSON)

Pass a single JSON object (via stdin or a file path):

Field	Required	Description
theme	Yes	Vibe name, e.g. "LOTR study", "Italian café"
audioDir	Yes	Path to folder containing .wav files (sorted by filename)
imagePath	Yes	Path to one background image (e.g. PNG)
outputDir	No	Where to write the MP4 (default: project Outputs/)
preset	No	"study" (default) or "epic" (slower zoom, stronger vignette)
outputFileName	No	Custom output filename; otherwise {theme}_{timestamp}.mp4
targetDurationMinutes	No	Target length in minutes (e.g. 60). Merged audio is looped and trimmed to this duration.
targetDurationSeconds	No	Target length in seconds (e.g. 3600). Overrides minutes if both are set.

When targetDurationMinutes or targetDurationSeconds is set, the merged audio is looped (repeated) and then trimmed to exactly that length, so you can hit a 45–90 minute (or any) target from shorter source WAVs.

Example (stdin):

echo '{"theme":"LOTR study","audioDir":"Audio_Parts","imagePath":"Images/tt.png","targetDurationMinutes":60}' | node Script/run-pipeline.js

Example (file):

node Script/run-pipeline.js /path/to/input.json

Output (JSON to stdout)

On success, the script prints a single JSON object:

• outputPath — path to the generated MP4
• durationSeconds — length of the video
• theme — echoed back
• title — template-based, e.g. "LOTR study | 1h 30m Study / Focus Music"
• description — short blurb (theme, duration, "for studying/focus")

On error, it prints { "error": "message" } and exits with a non-zero code.

Running from N8N

1. Trigger (schedule or webhook) → set variables: theme, audioDir, imagePath (e.g. from a "save files" step).
2. Execute Command (or Code node): run node Script/run-pipeline.js with the input JSON (e.g. write to a temp file and pass its path, or pipe JSON into stdin).
3. Parse the JSON output.
4. (Optional) Use an AI node (OpenAI / N8N AI) with theme, durationSeconds, and outputPath to generate a richer title and description; use AI output when present, otherwise keep the script's title and description.
5. Use outputPath, title, and description in a YouTube upload (or file move / notification) node.

Requirements: Node.js 14+ and FFmpeg (and ffprobe) on the PATH. No npm dependencies; the script uses only Node built-ins.

Presets

• study (default): Calm, slow Ken Burns zoom, light vignette, subtle film grain.
• epic: Slower zoom, stronger vignette for a "big" feel (e.g. LOTR-style).

You can extend presets in Script/presets.js.