01AUTHOR REPORTED
RTX 4090 · ComfyUI port
10 seconds of 1280 × 704 video in 85 seconds; 832 × 480 in 34 seconds. Warm run, two 5-second segments.
Port author’s README · checked 2026-10-29
THE INDEPENDENT FIELD GUIDE 01 / 04
Hy-Nomas
A world you can
move through.
Tencent Hunyuan’s interactive video world model, made approachable. Find the files, set up ComfyUI, and understand what the hardware reports actually say.
UNOFFICIAL RESOURCE · NOT AFFILIATED WITH TENCENT
OPEN MODEL✳TEXT + CAMERA✳LOCAL SETUP✳INDEPENDENT SOURCES
01THE ESSENTIALS
Three files. One place to start.
The ComfyUI port uses two model files and a text encoder. Put each in its corresponding folder.
| File | Destination | Size | Source |
|---|---|---|---|
hy_nomas_distil_int8.safetensors | models/diffusion_models | 9.06 GB | Port model files |
hy_nomas_pixelumm_fp8.safetensors | models/vae | 9.13 GB | Port model files |
qwen_2.5_vl_7b_fp8_scaled.safetensors | models/text_encoders | 9.38 GB | Comfy-Org |
Model files total approximately 27.6 GB; allow extra space for ComfyUI, cache and output. File sizes use decimal GB. Verified 2026-10-29.
See the complete installation guide →↳CAPABILITIES
What Hy-Nomas produces.
Hy-Nomas generates a world in segments. Each new segment can continue from the model’s history, with text and camera movement guiding what happens next.
| Property | Port examples and controls |
|---|---|
| Resolution | 1280 × 704 or 832 × 480 |
| Frame rate | 24 fps |
| Segments | The published examples use 5-second segments, chained into a 20-second world. |
| Steering | Text for each segment, recorded camera paths, a typed key timeline, or keyboard-driven turns. |
| Start from an image | The port encodes a still into the initial world history; this is a port feature. |
The port’s example world prompts describe the published clips. They are prompt JSON files; open a separate workflow for the ComfyUI graph. Longer worlds require more resources and have not been tested by this site.
Source: Port samples and examples ↗ · checked 2026-10-29
02HARDWARE NOTES
What can your machine handle?
These figures come from the port author, not our own measurements. A simulated memory limit is not a test on a different GPU.
02AUTHOR REPORTED
12 / 16 / 24 GB · simulated
93 / 91 / 89 seconds respectively for 10 seconds at 1280 × 704. Simulated on an RTX 4090 by holding back the rest of its memory, before the last kernel fusions. Not measured on a 12, 16 or 24 GB card.
Port author’s README · checked 2026-10-29
03AUTHOR REPORTED
RTX 4090 · 20-second world
About 3 minutes 15 seconds at 1280 × 704, including 91 seconds of decoding. Starting from an image adds 2–6 seconds, according to the port author.
Port author’s README · checked 2026-10-29
03GETTING STARTED
From files to first world.
01
Install the nodes
Find “Hy-Nomas (world model)” in ComfyUI-Manager, or clone the third-party port.
02
Place the files
Download the DiT, decoder and text encoder into the folders listed above.
03
Open a workflow
Drag the port’s per-segment example into ComfyUI and follow its included notes.
04PRIMARY SOURCES
Go straight to the source.
05QUICK ANSWERS
Hy-Nomas: questions answered.
A concise guide to the model, setup, hardware and licensing, with sources for each answer.
What is Hy-Nomas?
Hy-Nomas is an interactive video world model released by Tencent Hunyuan. It generates a world in successive segments that can be guided with text and camera movement.
Source: Tencent model card · checked 2026-10-29
Can I run Hy-Nomas in ComfyUI?
Yes. A third-party ComfyUI port provides nodes and example workflows. Install the node pack, download the DiT, PixelUMM decoder and text encoder, then open a workflow. The port is a work in progress; this site has not tested it locally.
Source: Port quick start · checked 2026-10-29 · ComfyUI setup guide →
How much VRAM does Hy-Nomas need?
The port author lists an NVIDIA GPU with 12 GB or more, but the 12 GB and 16 GB results were simulated by limiting available memory on an RTX 4090. They are not measurements on those cards. System RAM and settings also matter.
Source: Port hardware requirements · checked 2026-10-29 · Hardware reports →
Can Hy-Nomas start from an image?
The ComfyUI port offers a Start From Image workflow. The port author explains that it encodes the image into the world’s history; this is a port feature, not a native image input trained into the underlying model.
Source: Port image workflow · checked 2026-10-29 · ComfyUI setup guide →
How do I control the camera in Hy-Nomas?
In the ComfyUI port, use Camera Recording to record a path with W/A/S/D and arrow keys, or use Camera Keys for a typed timeline such as "W 2, W+Left 1.5, idle 0.5". A camera path exported from Tencent’s demo can be pasted into Camera Recording under { } JSON. Drive mode lets you steer the world one turn at a time.
Source: Port camera controls · checked 2026-10-29 · Camera control guide →
Can I use Hy-Nomas commercially?
Check the licenses for the specific files and intended use. The port’s NOTICE says the PixelUMM decoder weights and their fine-tunes are subject to the NVIDIA One-Way Noncommercial License, which limits use to non-commercial research and evaluation.
Source: PixelUMM license text · checked 2026-10-29 · Licensing information →