메뉴
Increase video frame rate by 2× or 4× using AI optical flow interpolation (RAFT). Produces smoother motion from low-fps footage. Requires a CUDA GPU with at least 2 GB free VRAM.
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 | frame-interpolator |
| description | Increase video frame rate by 2× or 4× using AI optical flow interpolation (RAFT). Produces smoother motion from low-fps footage. Requires a CUDA GPU with at least 2 GB free VRAM. |
| metadata | {"openclaw":{"emoji":"🎞️","requires":{"bins":["uv","ffmpeg"]}}} |
Increases video frame rate using RAFT optical flow to synthesise intermediate frames. 24 fps → 48/96 fps, or any source fps × 2/4.
The skill directory (where this SKILL.md lives) is referred to as $SKILL_DIR below.
GPU required. Needs a CUDA GPU with ≥ 2 GB free VRAM. If the LLM context is occupying VRAM, pause it before running.
Use this skill when the user wants to:
cd "$SKILL_DIR" && uv sync
RAFT model weights are downloaded automatically from torchvision on first run.
Before I interpolate the video(s), I need to know:
🎞️ Multiplier
- 2 — double the frame rate (e.g. 24 → 48 fps) [default]
- 4 — quadruple the frame rate (e.g. 24 → 96 fps)
🧠 Model
- raft_large — higher quality, more VRAM [default]
- raft_small — faster, uses less VRAM
📁 Input / output directories (default: ./input and ./output)
Wait for user response before proceeding.
Write or update $SKILL_DIR/config.json based on the user's choices.
cd "$SKILL_DIR" && uv run python scripts/interpolate.py --config config.json
Tell the user the output file paths, original vs new frame rate, and frame count.
| Key | Values | Default | Description |
|---|---|---|---|
input_dir | path | ./input | Folder containing input videos |
output_dir | path | ./output | Destination folder |
multiplier | 2, 4 | 2 | Frame rate multiplier |
model | raft_large, raft_small | raft_large | RAFT model variant |
# Default: 2× interpolation with raft_large
cd "$SKILL_DIR" && uv run python scripts/interpolate.py
# 4× frame rate
cd "$SKILL_DIR" && uv run python scripts/interpolate.py --multiplier 4
# Faster, lower VRAM usage
cd "$SKILL_DIR" && uv run python scripts/interpolate.py --model raft_small
# Custom directories
cd "$SKILL_DIR" && uv run python scripts/interpolate.py \
--input /path/to/clips --output /path/to/smooth
Each input video produces one output .mp4 with the same filename in output_dir.
Frame rate is multiplied; duration stays the same.