Sora 3 Unofficial

Local AI video + audio · No inference API

Real Wan 2.2 TI2V-5B video generation with optional Stable Audio 3 Small-SFX audio. Unofficial project; not affiliated with OpenAI.

Video prompt

Optional starting image (image-to-video)

Target audio duration (seconds)

5 10

Inference steps

10 50

Seed

Generate and merge Stable Audio 3 sound effects

Audio prompt

Generated MP4

Status

Requirements: a CUDA GPU with substantial VRAM and disk space. Stable Audio 3 is gated; accept its model terms on Hugging Face first. No API keys are used. The first launch downloads large model weights. This app has not been GPU-tested in this environment.