HunyuanVideo-I2V
HunyuanVideo-I2V extends the original HunyuanVideo model with reference-image conditioning. It incorporates image tokens into the video generation process.
Model design and access
The decoder-only multimodal text encoder processes the image and caption together. Tencent updated the released weights on 7 March 2025 to address first-frame identity changes. Use the corrected repository revision when reproducing the official workflow.HunyuanVideo-I2V official model documentation
Practical use
Practical guidance. Inspect the first frame and later frames separately. A matching starting image does not guarantee that facial features, product details or object proportions remain stable during motion.
Specifications & access
| Detail | Value and conditions |
|---|---|
| Access | Downloadable |
| License | Tencent Hunyuan Community License; custom terms applyHunyuanVideo-I2V official model documentation |
| Reference workflow | 720p image-to-videoHunyuanVideo-I2V official model documentation |
| Adaptation | LoRA training code for custom effectsHunyuanVideo-I2V official model documentation |
Related models
- HunyuanVideo
Tencent · HunyuanVideo
HunyuanVideo is the original text-to-video release published in December 2024.
Launch date Downloadable - HunyuanVideo-1.5
Tencent · HunyuanVideo
A text-to-video and image-to-video model combining an 8.3B diffusion transformer with a causal 3D VAE.
Launch date not confirmedDownloadable