Wan2GP is a web interface for video generation with Wan 2.2, HunyuanVideo and LTX Video, built to keep large models within limited video memory. It moves parts of a model into system RAM and picks a memory profile for your card, so large models also run on cards with less memory, only more slowly. On gpu.nz the template runs Wan2GP on a rented card of the RTX 40 generation or newer: you need no GPU of your own, and the interface opens through a protected link.
In short
Wan2GP runs on cards with 16 GB or more. The most affordable suitable card right now is the RTX A4000 16 GB from 13 ₽ per hour, billed by the minute.
Updated 13 October 2026
The template needs 16 GB and the Ada Lovelace architecture or newer (RTX 40, RTX 50, L40S, H100): RTX 30-series and A100 cards are greyed out for it in the rent sheet. Wan 2.2 5B and Wan 2.1 1.3B fit in 16–24 GB, for example on an RTX 4090. For Wan 2.2 14B and HunyuanVideo at good quality take 48 GB, for example an L40S: 14B also runs on 24 GB with offloading, but noticeably slower.
Prices and availability come from the live catalog and update automatically.
Wan2GP lists all its models in one place, so the easiest comparison is one prompt and one short clip per model. Wan 2.2 5B needs less memory, while Wan 2.2 14B is larger and heavier but usually gives the better result. HunyuanVideo and LTX Video are in the same list: try them on your own scenes and judge by the output.
| Model | Memory it fits | When to pick it |
|---|---|---|
| Wan 2.2 5B | 16–24 GB | drafts and short clips |
| Wan 2.2 14B | 48 GB for good quality | final clips |
| HunyuanVideo | 48 GB for good quality | a side-by-side check against Wan on one prompt |
For a video graph of your own, see video generation on a rented GPU, which covers Wan and HunyuanVideo in ComfyUI.
Generation time depends on the model, the resolution and the clip length, so start with a short test at a low resolution. Take the disk with some room: the weights of large models run to tens of gigabytes, and the template recommends 128 GB.