Ovi: Open Source AI Video & Audio Generator Like Sora 2 and Veo 3

File Information

Property	Details
Name	Ovi
Version	Latest commit on GitHub
File Type	Git repository (zip available)
License	Open Source (Apache 2.0)
Repository	GitHub Repository
Platform	Windows, Linux, macOS
Required Python	3.10+
Dependencies	PyTorch, torchvision, torchaudio, flash_attn, others

Description

Ovi is a groundbreaking open-source AI model developed by Character AI that revolutionizes the way creators generate audiovisual content. Ovi can simultaneously produce synchronized video and audio directly from text or a combination of text and images. Inspired by advanced models like Veo 3 and Sora 2, it empowers users to bring their ideas to life with stunning cinematic results, without the need for expensive software or high-end hardware.

With Ovi, creators can produce 5-second high-quality videos with audio at 24 frames per second, supporting multiple aspect ratios including 9:16, 16:9, and 1:1. This makes it perfect for social media clips, marketing content, and even educational short videos. The model’s flexible input system allows you to simply provide a text prompt describing the scene, or enhance it further by adding an image reference, giving you full creative control over the generated content.

One of Ovi’s most impressive features is its ability to generate synchronized audio and video seamlessly. Using special tags like <S> for speech and <AUDCAP> for audio descriptions, users can define exactly what dialogue or sound effects appear in the video. This ensures that every generated clip has perfectly timed narration, effects, or background sounds, opening endless possibilities for storytelling, music videos, or even animated scenes.

Ovi is also designed to be highly accessible. The model is open-source, meaning developers can modify it, integrate it into their own projects, or experiment with custom enhancements. Its performance is optimized for modern GPUs, but it also works with slightly older hardware, making it approachable for both hobbyists and professional creators.

For creators who love experimenting, Ovi supports multiple modes including text-to-video, image-to-video, and hybrid workflows. You can generate clips in bulk, test variations, and even fine-tune the output using configuration files. Multi-GPU setups are supported for faster generation, while single GPU usage is straightforward for beginners.

Usage of Ovi AI Video Generator

Whether you’re a filmmaker, content creator, or AI enthusiast, Ovi provides a powerful, flexible, and fully open-source tool to produce high-quality audiovisual content. With its combination of ease-of-use, creative freedom, and advanced AI capabilities, Ovi truly sets itself apart as one of the most exciting video generation tools available today.

You can easily find some of the video Generations from Sora 2 & veo3 alternative Ovi on their official github repository

How Ovi was Developed??

Ovi is built with contributions and ideas from some of the best open-source video and audio generation projects. Its video branch is initialized from Wan2.2, ensuring cutting-edge capabilities for text-to-video generation. The audio encoder and decoder components are borrowed from MMAudio, which allows Ovi to synchronize audio with video seamlessly.

Features of Ovi AI Video Generator

Feature	Description
Video+Audio Generation	Create synchronized video and audio clips in one go
Flexible Input	Supports text-only or text+image prompts
Multi-Aspect Ratios	9:16, 16:9, 1:1 and more
High FPS	Generates 5-second clips at 24 FPS
Customizable Prompts	Use <S> tags for speech and <AUDCAP> tags for audio descriptions
Multi-GPU Support	Accelerated generation with multiple GPUs
Open-Source	Fully modifiable and extensible
Easy Configuration	YAML files to control generation quality, audio/video balance, and output directory
Example Prompts	Ready-to-use CSV examples for quick start
Flexible Memory Usage	Supports fp8 quantization and CPU offloading for lower VRAM GPUs

Screenshots

System Requirements

Component	Minimum	Recommended
GPU	32 GB VRAM for standard model, 24 GB for fp8 quantized	RTX 40XX, 30XX, 20XX, GTX 10XX, or equivalent
CPU	8 cores	12+ cores for multi-GPU setups
RAM	16 GB	32+ GB
Disk Space	10 GB	50+ GB for models, videos, and caches
OS	Windows, Linux, macOS	Same
Python	3.10+	Same

How to Install Ovi: An Open Source Alternative to Sora 2 & Veo3 Locally

Step-by-Step Installation

1. Clone the repository

You can clone the repository from command below or download it manually from download section of the page.

git clone https://github.com/character-ai/Ovi.git
cd Ovi

2. Create and activate a virtual environment

virtualenv ovi-env
source ovi-env/bin/activate

3. Install PyTorch

pip install torch==2.5.1 torchvision==0.20.1 torchaudio==2.5.1

4. Install dependencies

pip install -r requirements.txt

5. Install Flash Attention

pip install flash_attn --no-build-isolation

Alternative method (if needed)

git clone https://github.com/Dao-AILab/flash-attention.git
cd flash-attention/hopper
python setup.py install
cd ../..

6. Download pre-trained weights

python3 download_weights.py
# Optional: custom directory
python3 download_weights.py --output-dir <custom_dir>

7. Run Ovi

python3 inference.py --config-file ovi/configs/inference/inference_fusion.yaml

For multi-GPU setups:

torchrun --nnodes 1 --nproc_per_node 8 inference.py --config-file ovi/configs/inference/inference_fusion.yaml

You can also launch the Gradio interface:

python3 gradio_app.py
# Optional flags: --cpu_offload, --use_image_gen, --fp8

Download Ovi: An Open Source Veo3 & Sora2 Alernative to generate AI video with audio for free

Tips & Advantages of Ovi AI Video + Audio Generator

• Generate videos directly from text prompts in seconds.
• Include synchronized speech or sound effects using special tags.
• Fine-tune video quality, denoising steps, and audio/video balance via config files.
• Share .lset style prompt files with embedded configurations for collaboration.
• Works efficiently on both older and newer GPUs, with fp8 quantization reducing VRAM usage.
• Open-source and fully modifiable, perfect for research or personal projects.

Ovi is an exceptional choice for anyone exploring AI-driven video creation. It offers the same capabilities as Sora 2 or Veo 3, but with full open-source flexibility, multi-GPU support, and customizable inputs for truly unique results. Whether you are making short clips for social media, experimental AI films, or storytelling videos, Ovi provides unmatched control and efficiency.