The-frizzy1/Hunyuan-Video-Low-VRAM-4GB

Hunyuan Video — Low VRAM 4GB Workflow (I2V + T2V)

By: The_frizzy1 Hardware target: RTX 3050 Laptop (4 GB VRAM) CivitAI: https://civitai.com/models/1312419/hunyuan-video-img2vid-low-vram-4gb-workflow YouTube: https://www.youtube.com/@the_frizzy1

Tencent Hunyuan Community License. Copyright © 2024 Tencent.

Tip: Wan2.1 performs better on low VRAM: https://huggingface.co/The-frizzy1/Wan21-GGUF-4GB-Workflow

What This Is

Semi-native ComfyUI workflow for Hunyuan Video on low-VRAM hardware. Includes I2V and T2V. Tested on RTX 3050 — works, but slow.

Notes

• —I2V encodes image into latent space for style — does not animate directly
• —T2V works well on low VRAM — check quantisation versions
• —Uses GGUF model + upscaler + RifleX
• —Low on VRAM? Replace sampler with Tiled KSampler
• —Keep all nodes updated — GGUF errors usually fixed by an update

Model Downloads

Fast Hunyuan (recommended): https://huggingface.co/city96/FastHunyuan-gguf/tree/main → Use fast-hunyuan-video-t2v-720p-Q5_K_M.gguf

Standard Hunyuan: https://huggingface.co/city96/HunyuanVideo-gguf/tree/main

VAE + CLIP: https://huggingface.co/calcuis/hunyuan-gguf/tree/main

• —hunyuan_video_vae_bf16.safetensors
• —llava_llama3_fp8_scaled.safetensors

Changelog