Sora (via Fal.ai)¶

OpenAI's Sora-2 model hosted on Fal.ai's infrastructure.

Quick Example¶

from tarash.tarash_gateway import generate_video
from tarash.tarash_gateway.models import VideoGenerationConfig, VideoGenerationRequest

config = VideoGenerationConfig(
    provider="fal",
    model="fal-ai/sora-2",
    api_key="YOUR_FAL_KEY",
)

# Text-to-video
request = VideoGenerationRequest(
    prompt="A futuristic cityscape at night with flying cars",
    duration_seconds=8,
    aspect_ratio="16:9",
)

response = generate_video(config, request)
print(response.video)

Supported Models¶

Model prefix	Duration	Image-to-Video	Notes
`fal-ai/sora-2`	`4s`, `8s`, `12s`	✅	Remix via `/video-to-video/remix`

Parameters¶

Parameter	Required	Supported	Notes
`prompt`	✅	✅
`duration_seconds`	—	✅	`4`, `8`, or `12`
`aspect_ratio`	—	✅	Passed through directly
`image_list` (reference)	—	✅	Image-to-video
`video`	—	✅	Video edit/remix input

Video Remix¶

To remix an existing video, use the /video-to-video/remix endpoint and pass the source video_id via extra_params.

# Sora-2 remix (requires fal-ai/sora-2/video-to-video/remix endpoint)
config = VideoGenerationConfig(
    provider="fal",
    model="fal-ai/sora-2/video-to-video/remix",
)

request = VideoGenerationRequest(
    prompt="Same scene in winter",
    extra_params={"video_id": "video_abc123"},
)