image-to-video

name	image-to-video
description	Animate image into video. Use for: bring photo to life, motion from still image
allowed-tools	Bash(curl *)
disable-model-invocation	true

Animate a static image into a video using fal.ai Wan 2.2 5B image-to-video model. The image must be a publicly accessible URL. Generation typically takes 1-3 minutes.

Ask the user for:

• IMAGE_URL: public URL of the image to animate (HTTPS URL required)
• PROMPT: description of the motion/animation to apply (e.g. "gentle waves in the background, leaves swaying")
• OUTPUT_FILE (optional): where to save — default: ~/Downloads/i2v_[timestamp].mp4

Then run the following steps:

Step 1 — Check API key:

(bash) if [ -z "$FAL_KEY" ]; then echo "ERROR: FAL_KEY not set." echo "Run /pocket-knife:setup to configure your API keys." exit 1 fi

IMAGE_URL="[PUBLIC_IMAGE_URL_HERE]" PROMPT="[ANIMATION_DESCRIPTION_HERE]" OUTPUT_FILE="${OUTPUT_FILE:-$HOME/Downloads/i2v_$(date +%Y%m%d_%H%M%S).mp4}"

MODEL="fal-ai/wan/v2.2-5b/image-to-video"

if [ -z "$IMAGE_URL" ]; then echo "ERROR: IMAGE_URL is required. Provide a public HTTPS URL to the source image." exit 1 fi (end bash)

Step 2 — Submit image-to-video job:

(bash)

Source: https://fal.ai/models/fal-ai/wan/v2.2-5b/image-to-video/api

Key difference from text-to-video: body includes both "prompt" AND "image_url"

image_url must be a publicly accessible HTTPS URL

Auth: "Authorization: Key $FAL_KEY" (NOT Bearer)

echo "Submitting image-to-video job..." SUBMIT=$(curl --fail-with-body -s
-X POST "https://queue.fal.run/${MODEL}"
-H "Authorization: Key $FAL_KEY"
-H "Content-Type: application/json"
-d "{"prompt": "${PROMPT}", "image_url": "${IMAGE_URL}"}")

if [ $? -ne 0 ]; then echo "ERROR: Failed to submit image-to-video job." echo "Response: $SUBMIT" echo "Check FAL_KEY in ~/.claude/.env and ensure image_url is a public HTTPS URL." exit 1 fi

REQUEST_ID=$(echo "$SUBMIT" | python3 -c "import sys,json; print(json.load(sys.stdin)['request_id'])" 2>/dev/null)

if [ -z "$REQUEST_ID" ]; then echo "ERROR: Could not extract request_id from response." echo "Response: $SUBMIT" exit 1 fi

echo "Job submitted. request_id: $REQUEST_ID" echo "Polling for completion (typically 1-3 minutes)..." (end bash)

Step 3 — Poll for completion:

(bash) MAX_WAIT=300 ELAPSED=0 POLL_INTERVAL=10

while [ $ELAPSED -lt $MAX_WAIT ]; do sleep $POLL_INTERVAL ELAPSED=$((ELAPSED + POLL_INTERVAL))

STATUS_RESP=$(curl -f -s
"https://queue.fal.run/${MODEL}/requests/${REQUEST_ID}/status"
-H "Authorization: Key $FAL_KEY")

if [ $? -ne 0 ]; then echo "WARNING: Status poll failed. Retrying... (${ELAPSED}/${MAX_WAIT}s)" continue fi

STATUS=$(echo "$STATUS_RESP" | python3 -c "import sys,json; print(json.load(sys.stdin).get('status',''))" 2>/dev/null) HAS_ERROR=$(echo "$STATUS_RESP" | python3 -c "import sys,json; d=json.load(sys.stdin); print('yes' if d.get('error') else 'no')" 2>/dev/null)

if [ "$HAS_ERROR" = "yes" ]; then echo "ERROR: Image-to-video generation failed." echo "$STATUS_RESP" exit 1 fi

echo "Status: $STATUS — ${ELAPSED}/${MAX_WAIT}s elapsed"

if [ "$STATUS" = "COMPLETED" ]; then echo "Video generation complete!" break fi done

if [ $ELAPSED -ge $MAX_WAIT ] && [ "$STATUS" != "COMPLETED" ]; then echo "ERROR: Image-to-video timed out after ${MAX_WAIT} seconds." echo "request_id: $REQUEST_ID" exit 1 fi (end bash)

Step 4 — Fetch result and download video:

(bash) RESULT=$(curl --fail-with-body -s
"https://queue.fal.run/${MODEL}/requests/${REQUEST_ID}/response"
-H "Authorization: Key $FAL_KEY")

if [ $? -ne 0 ]; then echo "ERROR: Failed to fetch result." exit 1 fi

VIDEO_URL=$(echo "$RESULT" | python3 -c "import sys,json; print(json.load(sys.stdin)['video']['url'])" 2>/dev/null)

if [ -z "$VIDEO_URL" ]; then echo "ERROR: Could not extract video URL from result." echo "Result: $RESULT" exit 1 fi

curl -f -s "$VIDEO_URL" --output "$OUTPUT_FILE"

if [ $? -ne 0 ] || [ ! -s "$OUTPUT_FILE" ]; then echo "ERROR: Video download failed or file is empty." rm -f "$OUTPUT_FILE" exit 1 fi

echo "Video saved to: $OUTPUT_FILE" (end bash)

Report to the user:

• The path where the video was saved
• The source image URL and animation prompt used
• Total time taken