HunyuanVideo 1.5 video model

HunyuanVideo 1.5 is out since 21st of November 2025. It's said to be the top open-source video generator so far.

It's built on DiT and aims to raise the bar for both ease of use and performance.

This version gives sharp visuals and smooth motion and still keeps things easy for devs and creators.

It’s light. With 8.3B parameters it runs on regular GPUs with just 14GB VRAM.

It makes 5 to 10 second videos in 480p or 720p, with support to boost them to 1080p for a more cinematic feel.

By mixing strong results with better hardware use it’s becoming the new standard for open-source tools.

Some users on Reddit are impressed with the new Hunyuan Video 1.5 model. It runs fast, sticks follows  prompts, works without LoRAs, and keeps character consistency better than older versions. Animations from still images look clean and it handles drawings well too. For some, this puts it ahead of Wan 2.2 in speed and prompt-following, though not always in quality.

Others argue Wan still offers sharper results, more control, and better features like ControlNets. A few users feel Hunyuan is just a solid fallback if you can't run Wan due to hardware limits. 

Some liked Hunyuan's animation style and noted better hand details than before, but blurry frames and lack of fine-tuning options remain a problem. There's talk that the model seems focused on Asian-style content and may not be fully general yet.

Users also shared tech setups and speeds. On a 3090, it takes around 2 minutes to make 2-second clips at 848x480. A 3060 user said 5-second clips took 10 minutes. 

There’s also a legal snag. The Hunyuan license blocks use in the EU, UK, and South Korea.

In December 2025 Hunyuan dropped a 480p step-distilled version for image-to-video. It makes clips in 8 or 12 steps, which they suggest using. On an RTX 4090 it cuts generation time by 75 percent making full videos in about 75 seconds on one card.