If you have a beefy GPU with at least 16gb vram you can use open source models like Wan 2.2 or LTX 2/2.3 for video generation, both takes text and image as input though for WAN you’d need to download the T2V - Text to Video for text input and for image input you’d need WAN I2V model, you can find plenty of guides online on it, all the models are available on huggingface or civitai and they run on ComfyUI where you basically have to connect nodes to use these models and loras
This is a really interesting topic, Has the software got to the point where it can also produce lifelike moans etc? or is it just the visual side of things?
WAN is only for video generation however LTX 2 and 2.3 both can generate audio with the video as well, but the quality isn’t that good you can check some previews on civitai gallery.
I installed WanGP on my machine, but I have to admit I haven’t used it much yet.
I still feel there must be something easier to use in terms of usability
My hardware is older, I would say: an AMD Ryzen 5 5600X, 32 GB RAM, and an AMD Radeon RX 6800 (RDNA2, 16 GB VRAM). I wanted to try out some tools and LLMs first before investing in something new.
I’m basically looking for something that can create videos longer than three seconds, which is currently the limit with (my) WanGP — ideally with a more comfortable UI
Since you have an AMD GPU I don’t think ComfyUI supports it on Windows yet, however I think it should work on Linux, also there are some fine-tunes and LoRAs for WAN that allows up to 30 seconds of video, if you’re into much more longer videos then LTX is the only way you can get those, as for the online tools I doubt any of those are uncensored.
Also I’ve only ever used ComfyUI for image and video generation so idk much about WANGP and AMD GPUs will be slower in generation compared to Nvidia GPUs so there’s that
