frame-interpolator

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"]}}}

tween — AI Frame Interpolation

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.

---

When to Use

Use this skill when the user wants to:

• Make slow-motion footage smoother
• Increase the frame rate of a video clip (2× or 4×)
• Remove judder from low-fps recordings

---

Setup (first run only)

cd "$SKILL_DIR" && uv sync

RAFT model weights are downloaded automatically from torchvision on first run.

---

Agent Workflow

1. Ask the user

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.

2. Edit config.json

Write or update $SKILL_DIR/config.json based on the user's choices.

3. Run

cd "$SKILL_DIR" && uv run python scripts/interpolate.py --config config.json

4. Report results

Tell the user the output file paths, original vs new frame rate, and frame count.

---

Config Reference

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

---

Common Invocations

# 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

---

Output

Each input video produces one output .mp4 with the same filename in output_dir. Frame rate is multiplied; duration stays the same.

---

Error Handling

• Insufficient VRAM → prints required vs available GB, tips to free VRAM, exits
• No CUDA GPU → clear error message, exits
• No videos in input_dir → clean message, exits
• Individual video errors → logged; other videos continue processing