Chetwin Low * 1 , Weimin Wang * † 1 , Calder Katyal 2
* Equal contribution, † Project Lead
1 Character AI, 2 Yale University
final_ovi_trailer.mp4
Ovi is a veo-3 like, video+audio generation model that simultaneously generates both video and audio content from text or text+image inputs.
- 🎬 Video+Audio Generation: Generate synchronized video and audio content simultaneously
- 🎵 High-Quality Audio Branch: We designed and pretrained our 5B audio branch from scratch using our high quality in-house audio datasets
- 📝 Flexible Input: Supports text-only or text+image conditioning
- ⏱️ 5-second Videos: Generates 5-second videos at 24 FPS, area of 720×720, at various aspect ratios (9:16, 16:9, 1:1, etc)
- 🎯 High-Resolution Support: Feel free to try 960×960 area (e.g., 720×1280, 704×1344, etc) - it could give outstanding results for both t2v and i2v! See examples below:
- 🎬 Create videos now on wavespeed.ai: https://wavespeed.ai/models/character-ai/ovi/image-to-video & https://wavespeed.ai/models/character-ai/ovi/text-to-video
- 🎬 Create videos now on HuggingFace: https://huggingface.co/spaces/akhaliq/Ovi
- 🔧 ComfyUI Integration (WIP): ComfyUI support is now available via ComfyUI-WanVideoWrapper, related PR.
- 🧠 Training Resolution: Our model was trained entirely under 720×720 resolution.
- 🚀 Upscaling Capability: Despite this, Ovi can generate naturally to higher resolutions such as 960×960 and variable-aspect videos (e.g., 1280×704, 1504×608, 1344×704) while maintaining temporal and spatial consistency.
An_older_man_with_a_full_grey_beard_and_long_grey__1280x720_104_4.mp4 |
A_concert_stage_glows_with_red_and_purple_lights.__1280x720_104_0.mp4 |
A_kitchen_scene_features_two_women._On_the_right.__704x1280_103_1.mp4 |
A_man_in_a_red_long-sleeved_shirt_and_dark_trouser_704x1280_104_3.mp4 |
The_scene_opens_on_a_dimly_lit_stage_where_three_m_704x1280_103_6.mp4 |
Two_men_are_shown_in_a_medium_close-up_shot_agains_704x1280_104_0.mp4 |
Two_women_stand_facing_each_other_in_what_appears__704x1280_103_0.mp4 |
Click the ⛶ button on any video to view full screen.
- Release research paper and website for demos
- Checkpoint of 11B model
- Inference Codes
- Text or Text+Image as input
- Gradio application code
- Multi-GPU inference with or without the support of sequence parallel
- fp8 weights and improved memory efficiency (credits to @rkfg)
- qint8 quantization thanks to @gluttony-10
- Improve efficiency of Sequence Parallel implementation
- Implement Sharded inference with FSDP
- Video creation example prompts and format
- Finetune model with higher resolution data, and RL for performance improvement.
- New features, such as longer video generation, reference voice condition
- Distilled model for faster inference
- Training scripts
We provide example prompts to help you get started with Ovi:
- Text-to-Audio-Video (T2AV):
example_prompts/gpt_examples_t2v.csv - Image-to-Audio-Video (I2AV):
example_prompts/gpt_examples_i2v.csv
Our prompts use special tags to control speech and audio:
- Speech:
<S>Your speech content here<E>- Text enclosed in these tags will be converted to speech - Audio Description:
<AUDCAP>Audio description here<ENDAUDCAP>- Describes the audio or sound effects present in the video
For easy prompt creation, try this approach:
- Take any example of the csv files from above
- Tell gpt to modify the speeches inclosed between all the pairs of
<S> <E>, based on a theme such asHuman fighting against AI - GPT will randomly modify all the speeches based on your requested theme.
- Use the modified prompt with Ovi!
Example: The theme "AI is taking over the world" produces speeches like:
<S>AI declares: humans obsolete now.<E><S>Machines rise; humans will fall.<E><S>We fight back with courage.<E>
# Clone the repository
git clone https://github.com/character-ai/Ovi.git
cd Ovi
# Create and activate virtual environment
virtualenv ovi-env
source ovi-env/bin/activate
# Install PyTorch first
pip install torch==2.6.0 torchvision torchaudio
# Install other dependencies
pip install -r requirements.txt
# Install Flash Attention
pip install flash_attn --no-build-isolationIf the above flash_attn installation fails, you can try the Flash Attention 3 method:
git clone https://github.com/Dao-AILab/flash-attention.git
cd flash-attention/hopper
python setup.py install
cd ../.. # Return to Ovi directoryTo download our main Ovi checkpoint, as well as T5 and vae decoder from Wan, and audio vae from MMAudio
# Default is downloaded to ./ckpts, and the inference yaml is set to ./ckpts so no change required
python3 download_weights.py
# For qint8 also ues python3 download_weights.py
OR
# Optional can specific --output-dir to download to a specific directory
# but if a custom directory is used, the inference yaml has to be updated with the custom directory
python3 download_weights.py --output-dir <custom_dir>
# Additionally, if you only have ~ 24Gb of GPU vram, please download the fp8 quantized version of the model, and follow the following instructions in sections below to run with fp8
wget -O "./ckpts/Ovi/model_fp8_e4m3fn.safetensors" "https://huggingface.co/rkfg/Ovi-fp8_quantized/resolve/main/model_fp8_e4m3fn.safetensors"
Ovi's behavior and output can be customized by modifying ovi/configs/inference/inference_fusion.yaml configuration file. The following parameters control generation quality, video resolution, and how text, image, and audio inputs are balanced:
# Output and Model Configuration
output_dir: "/path/to/save/your/videos" # Directory to save generated videos
ckpt_dir: "/path/to/your/ckpts/dir" # Path to model checkpoints
# Generation Quality Settings
num_steps: 50 # Number of denoising steps. Lower (30-40) = faster generation
solver_name: "unipc" # Sampling algorithm for denoising process
shift: 5.0 # Timestep shift factor for sampling scheduler
seed: 100 # Random seed for reproducible results
# Guidance Strength Control
audio_guidance_scale: 3.0 # Strength of audio conditioning. Higher = better audio-text sync
video_guidance_scale: 4.0 # Strength of video conditioning. Higher = better video-text adherence
slg_layer: 11 # Layer for applying SLG (Skip Layer Guidance) technique - feel free to try different layers!
# Multi-GPU and Performance
sp_size: 1 # Sequence parallelism size. Set equal to number of GPUs used
cpu_offload: False # CPU offload, will largely reduce peak GPU VRAM but increase end to end runtime by ~20 seconds
fp8: False # load fp8 version of model, will have quality degradation and will not have speed up in inference time as it still uses bf16 matmuls, but can be paired with cpu_offload=True, to run model with 24Gb of GPU vram
# Input Configuration
text_prompt: "/path/to/csv" or "your prompt here" # Text prompt OR path to CSV/TSV file with prompts
mode: ['i2v', 't2v', 't2i2v'] # Generate t2v, i2v or t2i2v; if t2i2v, it will use flux krea to generate starting image and then will follow with i2v
video_frame_height_width: [512, 992] # Video dimensions [height, width] for T2V mode only
each_example_n_times: 1 # Number of times to generate each prompt
# Quality Control (Negative Prompts)
video_negative_prompt: "jitter, bad hands, blur, distortion" # Artifacts to avoid in video
audio_negative_prompt: "robotic, muffled, echo, distorted" # Artifacts to avoid in audiopython3 inference.py --config-file ovi/configs/inference/inference_fusion.yamlUse this for single GPU setups. The text_prompt can be a single string or path to a CSV file.
torchrun --nnodes 1 --nproc_per_node 8 inference.py --config-file ovi/configs/inference/inference_fusion.yamlUse this to run samples in parallel across multiple GPUs for faster processing.
Below are approximate GPU memory requirements for different configurations. Sequence parallel implementation will be optimized in the future. All End-to-End time calculated based on a 121 frame, 720x720 video, using 50 denoising steps. Minimum GPU vram requirement to run our model is 32Gb, fp8 parameters is currently supported, reducing peak VRAM usage to 24Gb with slight quality degradation.
| Sequence Parallel Size | FlashAttention-3 Enabled | CPU Offload | With Image Gen Model | Peak VRAM Required | End-to-End Time |
|---|---|---|---|---|---|
| 1 | Yes | No | No | ~80 GB | ~83s |
| 1 | No | No | No | ~80 GB | ~96s |
| 1 | Yes | Yes | No | ~80 GB | ~105s |
| 1 | No | Yes | No | ~32 GB | ~118s |
| 1 | Yes | Yes | Yes | ~32 GB | ~140s |
| 4 | Yes | No | No | ~80 GB | ~55s |
| 8 | Yes | No | No | ~80 GB | ~40s |
We provide a simple script to run our model in a gradio UI. It uses the ckpt_dir in ovi/configs/inference/inference_fusion.yaml to initialize the model
python3 gradio_app.py
OR
# To enable cpu offload to save GPU VRAM, will slow down end to end inference by ~20 seconds
python3 gradio_app.py --cpu_offload
OR
# To enable an additional image generation model to generate first frames for I2V, cpu_offload is automatically enabled if image generation model is enabled
python3 gradio_app.py --use_image_gen
OR
# To run model with 24Gb GPU vram. No need to download additional models.
python3 gradio_app.py --cpu_offload --qint8
# To run model with 24Gb GPU vram
python3 gradio_app.py --cpu_offload --fp8
We would like to thank the following projects:
- Wan2.2: Our video branch is initialized from the Wan2.2 repository
- MMAudio: We reused MMAudio's audio vae.
We welcome all types of collaboration! Whether you have feedback, want to contribute, or have any questions, please feel free to reach out.
Contact: Weimin Wang for any issues or feedback.
If Ovi is helpful, please help to ⭐ the repo.
If you find this project useful for your research, please consider citing our paper.
@misc{low2025ovitwinbackbonecrossmodal,
title={Ovi: Twin Backbone Cross-Modal Fusion for Audio-Video Generation},
author={Chetwin Low and Weimin Wang and Calder Katyal},
year={2025},
eprint={2510.01284},
archivePrefix={arXiv},
primaryClass={cs.MM},
url={https://arxiv.org/abs/2510.01284},
}
