process-pipeline

name

process-pipeline

description

Runs the StickerFramework image processing pipeline on raw sticker images. Use when the user wants to process raw PNG images into platform-ready outputs (background removal, die-cut outline, resize, format conversion), generate animated TGS or video WEBM variants, or produce print sheets and distribution ZIPs. Don't use for generating new images via DALL-E, uploading to LINE, or publishing to Telegram.

Skill: process-pipeline

Purpose

Execute the image processing pipeline on an existing pack's raw images, producing platform-ready outputs in packs/<pack_id>/final/.

Prerequisites

• Python 3.10+ with dependencies installed (pip install -r requirements.txt).
• Raw source images exist in packs/<pack_id>/raw/ (PNG format).
• packs/<pack_id>/pack_config.py exists and is valid.

Steps

Step 1 — Resolve pack_id and paths

Determine pack_id from context:

• If the user specifies a pack name, use it directly.
• If a pack_config.py path is given, extract pack_id from it.
• Otherwise, ask the user which pack to process.

Verify the raw directory exists:

python3 .opencode/skills/process-pipeline/scripts/check_inputs.py "<pack_id>"

If the script reports missing files, inform the user and stop.

Step 2 — Choose processing mode

Determine which outputs are needed (infer from context or ask the user):

Mode	When to use	Step
Static only (default)	LINE, WhatsApp, Telegram static	Step 3
Skip background removal	Images already have transparent BGs	Step 3 with --skip-bg
LINE cover images	Always required for LINE upload	Step 3b (always run after static)
Animated TGS	Telegram animated stickers	Step 3c (separate call)
Video WEBM	Telegram video stickers	Step 3d (separate call)

Important: --telegram-animated and --telegram-video must each be run as separate run_pipeline.py invocations. Do not combine them with the static run.

Step 3 — Run static processing pipeline

Execute from the repo root:

python3 scripts/run_pipeline.py \
    --pack packs/<pack_id>/pack_config.py \
    --process-only \
    --input packs/<pack_id>/raw

If --skip-bg is needed:

python3 scripts/run_pipeline.py \
    --pack packs/<pack_id>/pack_config.py \
    --process-only \
    --input packs/<pack_id>/raw \
    --skip-bg

Step 3b — Generate LINE cover images (line_main / line_tab)

run_pipeline.py does not auto-generate line_main/ or line_tab/. Generate them from the first sticker after static processing completes:

python3 - <<'EOF'
from PIL import Image
import pathlib, shutil

pack_id = "<pack_id>"
src = sorted(pathlib.Path(f"packs/{pack_id}/final/line").glob("*.png"))[0]

# Main image: 240x240
out_main = pathlib.Path(f"packs/{pack_id}/final/line_main")
out_main.mkdir(parents=True, exist_ok=True)
img = Image.open(src).convert("RGBA")
img.thumbnail((240, 240), Image.LANCZOS)
canvas = Image.new("RGBA", (240, 240), (0, 0, 0, 0))
canvas.paste(img, ((240 - img.width) // 2, (240 - img.height) // 2))
canvas.save(out_main / "main.png")

# Tab icon: 96x74
out_tab = pathlib.Path(f"packs/{pack_id}/final/line_tab")
out_tab.mkdir(parents=True, exist_ok=True)
img2 = Image.open(src).convert("RGBA")
img2.thumbnail((96, 74), Image.LANCZOS)
canvas2 = Image.new("RGBA", (96, 74), (0, 0, 0, 0))
canvas2.paste(img2, ((96 - img2.width) // 2, (74 - img2.height) // 2))
canvas2.save(out_tab / "tab.png")

print("Done: line_main/main.png and line_tab/tab.png")
EOF

Step 3c — Generate animated TGS (if requested)

Animated TGS must be run as a separate pipeline call — do not append --telegram-animated to the static pipeline call above:

python3 scripts/run_pipeline.py \
    --pack packs/<pack_id>/pack_config.py \
    --process-only \
    --input packs/<pack_id>/raw \
    --telegram-animated \
    --animation-preset bounce

Available presets: bounce, wiggle, pulse, spin, shake, float, pop_in, slide_in, tada, heartbeat. Prefer bounce or pulse to stay under the 64 KB TGS limit.

Step 3d — Generate video WEBM (if requested)

Similarly, video WEBM is a separate pipeline call:

python3 scripts/run_pipeline.py \
    --pack packs/<pack_id>/pack_config.py \
    --process-only \
    --input packs/<pack_id>/raw \
    --telegram-video

Step 4 — Verify outputs

Run the output checker to confirm all expected platform directories are populated:

python3 .opencode/skills/process-pipeline/scripts/check_outputs.py "<pack_id>"

The script prints a per-platform summary table. Report any missing outputs to the user.

Step 5 — Report results

Summarise what was produced:

Platform	Directory	File count
LINE stickers	packs/<pack_id>/final/line/	<n>
LINE main image	packs/<pack_id>/final/line_main/	1
LINE tab icon	packs/<pack_id>/final/line_tab/	1
WhatsApp	packs/<pack_id>/final/whatsapp/	<n>
Telegram	packs/<pack_id>/final/telegram/	<n>
iMessage	packs/<pack_id>/final/imessage_large/	<n>
Print	packs/<pack_id>/final/print_etsy/	<n>

Inform the user of the next steps. See references/next-steps.md for platform-specific follow-up commands.

Error Handling

Error	Resolution
Input directory not found	Confirm raw images are in packs/<pack_id>/raw/. Run generation first if needed.
Background removal failed for an image	Add --skip-bg flag to bypass flood-fill and use rembg ML fallback.
ValueError: Unsupported format: TGS	Do not include telegram_animated in the platforms list in pack_config.py. Run TGS via Step 3c instead.
Animated TGS file exceeds 64 KB limit	Switch to a simpler preset (pulse or bounce) — avoid tada or heartbeat.
Video WEBM exceeds 256 KB limit	Reduce animation duration in pack_config.py sticker animation.duration_ms.
FFmpeg not found	Install FFmpeg: brew install ffmpeg (macOS) or apt install ffmpeg (Linux).
line_main/ or line_tab/ missing	Run Step 3b — these are not auto-generated by run_pipeline.py.