Skit Thumbnail Generator

Location

Script: /path/to/scripts/skit_thumbnail.py

Usage

cd "/path/to/project"
python scripts/skit_thumbnail.py path/to/clip.mp4 --title "Title Text" --subtitle "Subtitle Text"

Arguments

Design

• Left stack: Red SKIT bar (rounded pill, left-aligned text) → 4px gap → Photo (rounded-rect, white border, drop shadow)
• Right column: Title (red) + Subtitle (white) at same font size (determined by longer string) → accent line → channel logo (scaled to fill)
• Background: Dark slate with diagonal stripe accents, red top/bottom bars
• Margins: 50px left, 25px all other sides, 28px gap between frame and text

Pitfalls (learned the hard way)

• ❌ Don't use full-bleed blurred frame as background — user rejected it, prefers clean dark slate with diagonal stripes
• ❌ Don't put SKIT label inside the photo — it goes ABOVE in a separate red bar (pill shape, left-aligned text)
• ❌ Don't add a second SKIT badge in the right zone — one is enough (the bar above the photo)
• ❌ Don't under-size text to leave empty space — title + subtitle should be the same large font, auto-determined by the longer string
• ❌ Don't use tiny corner logos — logo must be large (200-440px) and centered in remaining space
• ❌ Don't redesign from scratch — user prefers iterative refinement of the existing polaroid layout
• ❌ Don't guess the title/subtitle — ask the user what text they want, or if they already told you, use exactly what they said (subtitle is often just "Overwatch" for Overwatch skits)
• ❌ Don't extract from a random frame without asking — always ask the user which frame they want, then use --frame-time to pin it (0 = first frame, or a specific second)
• ✅ SKIT bar text at 45px (1.5x of initial try), left-aligned with 15px padding
• ✅ Title & subtitle share ONE font size via calc_font_size (binary search on longer string)
• ✅ Everything auto-fits within margins: 50px left, 25px elsewhere

YouTube upload after thumbnail

After the user picks a thumbnail, upload the video to the channel:

cd "/path/to/scripts-upload/scripts"
python yt_upload.py "path/to/video.mp4" \
  --title "SkitName In Overwatch | My Channel" \
  --thumbnail "path/to/thumb.png" \
  --private --schedule "YYYY-MM-DDT14:00:00Z"

Title convention (channel skits): "[Skit Name] In Overwatch | My Channel" — always this format for skit videos. The subtitle (e.g. "Overwatch") goes in the thumbnail only, not the YouTube title.

Schedule: User's timezone is IDT (UTC+3). 5pm IDT = 14:00 UTC. Ask the user what time to publish.

Workflow patterns

Quick single thumbnail:

python scripts/skit_thumbnail.py clip.mp4 --title "Title" --subtitle "Subtitle"

Picks a random mid-clip frame. Re-run if user doesn't like the result.

Batch comparison (user-picky workflow): Generate 3 thumbnails at different timestamps, show them inline, let the user pick.

1. Check video duration: ffprobe -v error -show_entries format=duration -of csv=p=0 clip.mp4
2. Pick 3 spread-out timestamps (e.g. early/mid/late thirds minus margins)
3. Generate each:

   python scripts/skit_thumbnail.py clip.mp4 --title "T" --subtitle "S" --frame-time 3 --output "skit_thumbnails/clip_t1.png"
   python scripts/skit_thumbnail.py clip.mp4 --title "T" --subtitle "S" --frame-time 12 --output "skit_thumbnails/clip_t2.png"
   python scripts/skit_thumbnail.py clip.mp4 --title "T" --subtitle "S" --frame-time 22 --output "skit_thumbnails/clip_t3.png"

4. Present all 3 inline via MEDIA: paths
5. If none works, iterate with 3 more timestamps
6. Once picked, that PNG is the final thumbnail for the upload step

Script maintenance notes

• --frame-time was added retrospectively (original script only had random extraction). If you patch the script, the function to modify is extract_frame(video_path, timestamp=None) — it replaces the old extract_random_frame(). Argparse entry in main() needs both the new flag in the parser and passing it through to make_skit_thumbnail().

Output

1280×720 PNG, saved to skit_thumbnails/<clipname>_skit_thumb.png.