Video Generation APIs

Thin Python wrappers around current video generation HTTP APIs and official SDKs. Model names and paths follow vendor docs as of early 2026 (verify in each provider’s console if a call fails).

Note: the examples below import this repository's wrapper modules directly.

Files

Setup

Use the project virtualenv (recommended):

cd /path/to/openvideogen
python3 -m venv .venv
source .venv/bin/activate
pip install openai python-dotenv requests google-genai

Install from PyPI:

pip install openvideogen

Copy .env.example to .env and fill in keys.

Usage

OpenAI

from openai_video import generate_video_openai

job = generate_video_openai(
    "A calico cat playing a piano on stage",
    seconds="8",
    size="1280x720",
    wait=False,  # True → create_and_poll until completed/failed
)
print(job.id, job.status)

Runway

from runway_video import generate_video_runway

result = generate_video_runway(
    "A serene landscape with mountains and a river at sunset",
    turbo=True,   # gen4_turbo vs gen4.5
    ratio="1280:720",
    duration=5,
)

Google Veo

from google_veo_video import generate_video_veo

video = generate_video_veo(
    "A bustling city street during a rainy night",
    aspect_ratio="16:9",
    duration_seconds=8,
    resolution="1080p",  # optional: 720p, 1080p, 4k
    # reference_image_paths=["./ref1.png", "./ref2.png"],  # optional, local files
)
print(getattr(video, "uri", video))

Luma

from luma_video import generate_video_luma

video = generate_video_luma(
    "A futuristic cityscape with flying cars",
    model="ray-2",
    aspect_ratio="16:9",
    resolution="720p",
    duration="5s",
)

BytePlus / Volcengine (Seedance)

Set BYTEDANCE_ARK_MODEL (or model= argument) to the model id or endpoint id from the Ark console.

from bytedance_video import create_bytedance_video_task, generate_video_bytedance

task = create_bytedance_video_task("A cartoon character dancing in a park")
# or poll until a terminal status:
# result = generate_video_bytedance("...", character_reference="https://...")

Meta

generate_video_meta is intentionally not implemented — there is no public Movie Gen REST API for arbitrary keys.

PyPI Release Flow

This repository is configured to publish to PyPI from GitHub Releases via trusted publishing.

1. Update version in pyproject.toml (for example 0.1.1).
2. Create a GitHub Release with a tag that matches the version exactly. Tags like v0.1.1 are also accepted.
3. The workflow in .github/workflows/pypi-publish.yml builds and validates distributions, then publishes to PyPI.

One-Time Setup

1. In PyPI, create project openvideogen (or publish once manually to create it).
2. In PyPI project settings, add a trusted publisher for this GitHub repo: owner/repo: this repository workflow: pypi-publish.yml environment: pypi
3. In GitHub repo settings, create environment pypi (optional protections recommended).