# List all active models
tail -n +2 ${SKILLS_DIR:-$HOME/.claude/skills}/c-kie-ai/models.jsonl | python3 -c "
import sys,json
for line in sys.stdin:
    m=json.loads(line)
    if not m.get('deprecated'):
        print(f\"{m['key']:<45} {m['api_model']}\")
"

# Filter by operation (e.g. image-to-video)
tail -n +2 ${SKILLS_DIR:-$HOME/.claude/skills}/c-kie-ai/models.jsonl | python3 -c "
import sys,json
for line in sys.stdin:
    m=json.loads(line)
    if 'image-to-video' in m['operation'] and not m.get('deprecated'):
        print(m['api_model'])
"

Provider A: kie.ai (Primary)

Base: https://api.kie.ai/api/v1 Auth: Authorization: Bearer $KIE_AI_API_KEY Key source: ~/.gsai/secrets.env → KIE_AI_API_KEY

source ~/.gsai/secrets.env

⚠️ kie.ai cannot fetch external URLs. All image_url fields must be base64 data URIs:

IMAGE_B64=$(python3 -c "import base64; print('data:image/png;base64,' + base64.b64encode(open('$IMG','rb').read()).decode())")

Create Task (standard endpoint)

RESULT=$(curl -s -X POST "https://api.kie.ai/api/v1/jobs/createTask" \
  -H "Authorization: Bearer $KIE_AI_API_KEY" \
  -H "Content-Type: application/json" \
  -d "{\"model\": \"$MODEL\", \"input\": $INPUT_JSON}")
TASK_ID=$(echo "$RESULT" | python3 -c "import sys,json; print(json.load(sys.stdin)['data']['taskId'])")

Poll Status

while true; do
  RESP=$(curl -s "https://api.kie.ai/api/v1/jobs/recordInfo?taskId=$TASK_ID" \
    -H "Authorization: Bearer $KIE_AI_API_KEY")
  STATE=$(echo "$RESP" | python3 -c "import sys,json; print(json.load(sys.stdin)['data']['state'])")
  case "$STATE" in
    success|completed|done) break ;;
    fail|failed|error) echo "FAILED: $STATE" && exit 1 ;;
  esac
  sleep 5
done
VIDEO_URL=$(echo "$RESP" | python3 -c "
import sys,json
d=json.load(sys.stdin)['data']
import json as j2
print(j2.loads(d['resultJson'])['resultUrls'][0])
")

Model Reference (kie.ai)

Key	`model` value	Use for	Required input fields
Hailuo t2v	`hailuo/02-text-to-video-pro`	MiniMax text→video	`prompt`
Hailuo i2v	`hailuo/02-image-to-video-pro`	MiniMax image→video	`prompt`, `image_url` (base64)
Sora-2 t2v	`sora-2-text-to-video`	Sora 2 text→video	`prompt`, `aspect_ratio`, `n_frames`
Sora-2 i2v	`sora-2-image-to-video`	Sora 2 image→video	`prompt`, `image_urls` (array, base64), `aspect_ratio`
Kling i2v	`kling/v2-5-turbo-image-to-video-pro`	Kling 2.5 i2v	`prompt`, `image_url`, `duration`, `aspect_ratio`
Kling avatar	`kling/ai-avatar-v1-pro`	Kling AI avatar	`prompt`, `image_url`
WAN t2v	`wan/2-6-text-to-video`	WAN 2.6	`prompt`, `aspect_ratio`
WAN i2v	`wan/2-6-image-to-video`	WAN 2.6 i2v	`prompt`, `image_url`, `aspect_ratio`
Seedance	`bytedance/seedance-1.5-pro`	Seedance 1.5	`prompt`, `image_url`, `aspect_ratio`
GPT Image	`gpt-image/1.5-text-to-image`	GPT image gen	`prompt`, `aspect_ratio`
GPT Image i2i	`gpt-image/1.5-image-to-image`	GPT image edit	`prompt`, `input_urls` (array, base64)
Imagen 4	`google/imagen4`	Imagen 4	`prompt`, `aspect_ratio`
Imagen 4 fast	`google/imagen4-fast`	Imagen 4 fast	`prompt`, `aspect_ratio`, `num_images`
Z-Image	`z-image`	Z-Image gen	`prompt`, `aspect_ratio`
Grok i2v	`grok-imagine/image-to-video`	Grok i2v	`prompt`, `image_url`
InfiniTalk	`infinitalk/from-audio`	Talking head from audio	`image_url`, `audio_url`
Seedream 4.5	`seedream/4.5-text-to-image`	Seedream image	`prompt`, `aspect_ratio`
MiniMax music	`music-01`	Text→music	`prompt`

Hailuo i2v — Full Example (labubu-shorts pipeline)

IMG_B64=$(python3 -c "import base64; print('data:image/png;base64,' + base64.b64encode(open('$PNG','rb').read()).decode())")

RESULT=$(curl -s -X POST "https://api.kie.ai/api/v1/jobs/createTask" \
  -H "Authorization: Bearer $KIE_AI_API_KEY" \
  -H "Content-Type: application/json" \
  -d "{\"model\": \"hailuo/02-image-to-video-pro\", \"input\": {\"prompt\": \"$PROMPT\", \"image_url\": \"$IMG_B64\"}}")

TASK_ID=$(echo "$RESULT" | python3 -c "import sys,json; print(json.load(sys.stdin)['data']['taskId'])")
echo "Task: $TASK_ID"

Sora-2 aspect_ratio mapping

UI aspect	Sora-2 value
16:9	`landscape`
9:16	`portrait`

n_frames (Sora-2 duration)

Duration	n_frames
~5s	`"10"`
~10s	`"15"`

Provider B: fal.ai (FLUX + Async Queue)

Use when: you need FLUX image gen, or prefer queue-based async for video. Base: https://fal.run (sync image) / https://queue.fal.run (async video) Auth: Authorization: Key $FAL_KEY Key source: ~/.gsai/secrets.env → FAL_KEY

source ~/.gsai/secrets.env

Image Generation (synchronous)

curl -s -X POST "https://fal.run/c-fal-ai/flux/dev" \
  -H "Authorization: Key $FAL_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "prompt": "your prompt here",
    "image_size": "landscape_16_9",
    "num_images": 1,
    "output_format": "png"
  }' | python3 -c "import sys,json; d=json.load(sys.stdin); print(d['images'][0]['url'])"

Aspect ratio enum (image_size): square_hd | landscape_16_9 | portrait_16_9 | landscape_4_3 | portrait_4_3

Image models: c-fal-ai/flux/dev (quality) · c-fal-ai/flux/schnell (fast)

Video Generation (async queue)

Pattern: submit → poll status_url → fetch response_url

Step 1 — Submit

RESULT=$(curl -s -X POST "https://queue.fal.run/$MODEL" \
  -H "Authorization: Key $FAL_KEY" \
  -H "Content-Type: application/json" \
  -d "$PAYLOAD")
STATUS_URL=$(echo "$RESULT" | python3 -c "import sys,json; print(json.load(sys.stdin)['status_url'])")
RESPONSE_URL=$(echo "$RESULT" | python3 -c "import sys,json; print(json.load(sys.stdin)['response_url'])")

Step 2 — Poll (every 5s until COMPLETED)

while true; do
  STATUS=$(curl -s "$STATUS_URL" -H "Authorization: Key $FAL_KEY" \
    | python3 -c "import sys,json; print(json.load(sys.stdin).get('status',''))")
  [ "$STATUS" = "COMPLETED" ] && break
  [ "$STATUS" = "FAILED" ] && echo "FAILED" && exit 1
  sleep 5
done

Step 3 — Get output URL

VIDEO_URL=$(curl -s "$RESPONSE_URL" -H "Authorization: Key $FAL_KEY" \
  | python3 -c "import sys,json; d=json.load(sys.stdin); print((d.get('video') or d.get('videos',[{}])[0]).get('url',''))")

fal.ai Model Reference

Key	fal endpoint	Use for
Image	`c-fal-ai/flux/dev`	Quality image gen
Image	`c-fal-ai/flux/schnell`	Fast image gen
Video t2v	`c-fal-ai/kling-video/v3/pro/text-to-video`	Kling 3
Video i2v	`c-fal-ai/kling-video/v3/pro/image-to-video`	Kling 3 i2v
Video t2v	`c-fal-ai/veo3`	Veo 3
Video t2v	`c-fal-ai/veo3.1`	Veo 3.1
Video i2v	`c-fal-ai/veo3.1/image-to-video`	Veo 3.1 i2v
Video t2v	`bytedance/seedance-2.0/text-to-video`	Seedance 2
Video i2v	`bytedance/seedance-2.0/image-to-video`	Seedance 2 i2v
Video t2v	`c-fal-ai/wan/v2.2-a14b/text-to-video`	WAN 2.2
Video i2v	`c-fal-ai/wan/v2.2-a14b/image-to-video`	WAN 2.2 i2v
Video t2v	`c-fal-ai/minimax/video-01-live`	MiniMax (Hailuo)

Common Payload Fields (video)

{
  "prompt": "...",
  "image_url": "https://... or data:image/png;base64,...",
  "aspect_ratio": "16:9",
  "duration": 5,
  "resolution": "720p"
}

Seedance duration range: 4–15s. Veo/Kling duration: 5 or 10s.

Veo 3 (kie.ai — different endpoint)

Veo 3 uses /veo/generate and /veo/record-info:

# Create
curl -s -X POST "https://api.kie.ai/api/v1/veo/generate" \
  -H "Authorization: Bearer $KIE_AI_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{"model":"veo3_fast","prompt":"...","generationType":"TEXT_2_VIDEO","aspectRatio":"16:9"}'

# Poll
curl -s "https://api.kie.ai/api/v1/veo/record-info?taskId=$TASK_ID" \
  -H "Authorization: Bearer $KIE_AI_API_KEY"
# successFlag: 0=generating, 1=success, 2/3=failed
# output: data.response.resultUrls[0]

Gotchas

kie.ai:

image_url must be base64 data URI — kie.ai cannot fetch external URLs
sora-2-* model names use hyphens, NOT slashes (unlike other models)
Polling state: success/completed/done → done; fail/failed/error → failed
Error code 402 = insufficient credits; 401 = bad API key
Full model registry: see models.jsonl in this skill directory
DISCONTINUED: google/nano-banana-pro returns 422 since 2026-02-20

fal.ai:

status_url and response_url use queue.fal.run domain — same Authorization: Key header required
Seedance 2.0 duration must be passed as a string ("5" not 5)
MiniMax Live model only accepts prompt + optional image_url — no aspect_ratio/duration overrides
Full model registry: see models.jsonl in this skill directory

Zoom Presets

Asset type	Zoom
AI whiteboard	`1.1x`
AI cinematic/photo	`1.15x`
App/screen/mobile	`none`
Static graphics	`1.15x`
Motion graphics	`none`

Output Paths

AI images: {brand_path}/creatives/brolls/images/{id}-{desc}.png
AI clips: {brand_path}/creatives/brolls/ai/{id}-{desc}.mp4

c-kie-ai

AI Media Generation — kie.ai + fal.ai

Step 0 — Always run first (model cache auto-sync)