README.md

image-to-video

Batch image-to-video pipeline with pluggable providers, cost estimation, and optional Face Lock biometric drift verification.

Install

pip install image-to-video                 # core (aiohttp, Pillow)
pip install "image-to-video[vision]"       # + Face Lock (numpy, opencv, mediapipe)

From source (editable):

pip install -e ./image-to-video

Console commands

Two entry points are installed:

Command	Module	Purpose
i2v	image_to_video.pipeline:main	Validate → submit → poll → download a batch
face-lock-pipeline	image_to_video.face_lock_bridge:main	Lock identity → animate → drift-verify → report

export KLING_API_KEY="sk-..."

# Estimate cost before spending (no API calls)
i2v --input inputs --dry-run --duration 10 --mode pro

# Run a batch (default provider: kling; also: goenhance)
i2v --input inputs --output outputs --provider kling

# Identity-locked end-to-end (requires the [vision] extra)
face-lock-pipeline --reference ref.png --subject "Aria" --input inputs --output outputs

Layout

image-to-video/
├── pyproject.toml
├── src/image_to_video/
│   ├── pipeline.py          # generic batch machinery + CLI
│   ├── providers.py         # VideoProvider interface + Kling/GoEnhance + cost
│   ├── face_lock_core.py    # biometric measurement / prompt / drift logic
│   └── face_lock_bridge.py  # lock → animate → verify → report
└── tests/                   # self-contained: no network, no cv2/mediapipe

Add a backend by subclassing VideoProvider in providers.py and registering it in PROVIDERS. Pricing constants and the GoEnhance endpoint/schema are approximate — verify against each backend's live API before production use.

Test

pip install -e "./image-to-video[dev]"
pytest image-to-video/tests -q