메뉴
Apply synthetic depth-of-field bokeh to photos using MiDaS monocular depth estimation. Makes phone photos look like they were shot with a fast prime lens. No depth sensor required — works from a single image.
Codex 또는 Claude로 설치 이 Prompt를 복사해 Codex, Claude 또는 다른 어시스턴트에 붙여 넣으면 Skill 페이지를 검토하고 설치를 진행할 수 있습니다.
SOC 직업 분류 기준
Schedule, create, and manage social media posts on Instagram and LinkedIn using the Buffer GraphQL API. Use when the user wants to publish content, schedule posts, check scheduled posts, list connected channels, or create content ideas on Buffer.
Research SEO opportunities through audits, keyword discovery, traffic validation, and site architecture analysis.
Maintains and applies brand identity across all content operations. Ensures every output aligns with the personal brand's voice, values, and visual identity.
MCP skill for canva. Provides 23 tools: upload-asset-from-url, upload-asset-from-file, resolve-shortlink, search-designs, get-design, get-design-pages, get-design-content, get-presenter-notes, import-design-from-url, merge-designs, export-design, get-export-formats, create-folder, move-item-to-folder, list-folder-items, search-folders, comment-on-design, list-comments, list-replies, reply-to-comment, generate-design, create-design-from-candidate, list-brand-kits
Plan ads, generate ad creative, and configure paid acquisition measurement. Use for paid campaign setup and tracking.
Generate growth ideas and low-cost marketing approaches. Use for campaign ideation, growth planning, and free-tool strategy exploration.
| name | bokeh-effect |
| description | Apply synthetic depth-of-field bokeh to photos using MiDaS monocular depth estimation. Makes phone photos look like they were shot with a fast prime lens. No depth sensor required — works from a single image. |
| metadata | {"openclaw":{"emoji":"📷","requires":{"bins":["uv"]}}} |
Estimates per-pixel depth from each image using Intel MiDaS, then applies variable lens blur: near objects stay sharp, far objects blur progressively.
The skill directory (where this SKILL.md lives) is referred to as $SKILL_DIR below.
Use this skill when the user wants to:
cd "$SKILL_DIR" && uv sync
MiDaS weights are downloaded from torch.hub on first use (~400 MB for DPT_Large).
Before I process the images, I need to know:
🌀 Blur strength (1–50, default: 15)
Stronger = more pronounced bokeh.
Subtle effect: 5–10 | Natural: 12–18 | Heavy: 20+
🧠 Model
- DPT_Large — most accurate depth, slower (default)
- DPT_Hybrid — good balance
- MiDaS_small — fastest, lower depth quality
⚙️ Device
- auto — GPU if available, else CPU (default)
- cpu — force CPU (~5s/image, fine for small batches)
📁 Input / output directories (default: ./input and ./output)
cd "$SKILL_DIR" && uv run python scripts/bokeh.py --config config.json
| Key | Values | Default | Description |
|---|---|---|---|
input_dir | path | ./input | Source image folder |
output_dir | path | ./output | Destination folder |
model | DPT_Large, DPT_Hybrid, MiDaS_small | DPT_Large | Depth model |
blur_strength | 1–50 | 15 | Maximum blur radius in pixels |
device | auto, cpu, cuda | auto | Inference device |
# Default settings
cd "$SKILL_DIR" && uv run python scripts/bokeh.py
# Stronger blur
cd "$SKILL_DIR" && uv run python scripts/bokeh.py --blur-strength 25
# Fast mode (lower quality depth)
cd "$SKILL_DIR" && uv run python scripts/bokeh.py --model MiDaS_small
# Force CPU
cd "$SKILL_DIR" && uv run python scripts/bokeh.py --device cpu
Processed images written to output_dir with the same filenames as input.