WorkflowListedExternal

HunyuanVideo 12GB VRAM Workflow

With the native comfy implementation of Hunyuan I have tweaked the workflow to work for 12GB VRAM cards. It looks like you can get at least to 73 frames and probably a bit more. It takes about 8 minutes for a 4070Ti...

Sourcecivitai.com/models/1048302

TaskText to videoFamilyHunyuanEstimated VRAMEstimate unavailableBgrade

Runtime profile

Access: External workflow
Status: Listed
Source: Civitai
Updated: Jul 20, 2026

Source description

Do make sure to update your comfy and to get the exact result above the guidance I put up to 10 (I reduced it for the base workflow as it caused some burning for some prompts).

As its not as complete as the wrapper node there is a few less features than that for now. But there is also no crazy special installations that you need to do.

Links to model downloads here:

AI-generated commentary

AI-generated explanation based on source and configuration details. Suggestions are clearly labeled.

Purpose

This ComfyUI workflow is for generating Hunyuan Video clips from a text prompt: it encodes the text and outputs a combined video, and it is described as tweaked for cards with 12GB of VRAM.

Flow

The graph encodes text, creates an empty Hunyuan latent video, samples it, decodes it with tiled VAE processing, and combines the result into a video.

Inputs and outputs

The description says the output can reach at least 73 frames, possibly more.
The listing reports about 8 minutes for a 4070Ti to run 20 steps.

Setup

Before using it, update ComfyUI, as the description instructs.
Suggestion · not verified
The native ComfyUI implementation is described as not requiring special installations.

Dependencies

Named model files are clip_l.safetensors, hunyuan_video_t2v_720p_bf16.safetensors, hunyuan_video_vae_bf16.safetensors, and llava_llama3_fp8_scaled.safetensors.
Named nodes include BasicGuider, BasicScheduler, CLIPTextEncode, DualCLIPLoader, EmptyHunyuanLatentVideo, FluxGuidance, KSamplerSelect, and ModelSamplingSD3.
It also names RandomNoise, SamplerCustomAdvanced, UNETLoader, VAEDecodeTiled, VAELoader, and VHS_VideoCombine.

Configuration

If you are following the described setup, set guidance up to 10; the base workflow uses lower guidance because some prompts caused burning.
Suggestion · not verified

Cautions

Before use, verify that the listed model files and nodes are available in your ComfyUI setup.
Suggestion · not verified
The description notes that this implementation has fewer features than the wrapper node.

Does this need to be edited?

Sources

1 source

CivitaiPrimary

civitai.com/models/1048302

Verified Jul 20

Source excerpts

1 excerpt

NoteJul 20civitai.com/models/1048302

Source context: 16736 downloads · Type Workflows · Base model Hunyuan Video