Hyvid Open-Source Text-to-Video Model - Free Deployment to Generate High-Quality Anime-Style Videos

Hyvid

Developed by calcuis

An anime-style adapter based on Tencent's Hunyuan Video model, providing high-quality text-to-video generation capabilities, specifically optimized for anime-style content creation.

Text-to-Video EnglishOpen Source License:MIT #Anime Video Generation #FP8 Efficient Inference #LoRA Fine-Tuning Adaptation

Downloads 1,392

Release Time : 12/26/2024

Model Overview

This model is an anime-style adapter for Tencent's HunyuanVideo model, optimized through LoRA technology and GGUF quantization to generate high-quality anime-style video content. It supports text-to-video generation tasks and is particularly suitable for anime character and scene creation.

Model Features

Anime Style Optimization

Specifically optimized for anime characters and scenes, capable of generating high-quality anime-style content.

GGUF Quantization Support

Provides GGUF quantized versions to reduce hardware requirements and improve operational efficiency.

FP8 Scaling Technology

Utilizes FP8 scaling technology to optimize model components, reducing resource usage while maintaining quality.

Fast Loading

Optimized workflow and quantized files significantly improve model loading speed.

Model Capabilities

Text-to-Video Generation

Anime-Style Content Creation

Character and Scene Design

Stylized Video Generation

Use Cases

Anime Creation

Anime Character Design

Generates anime-style character images based on text descriptions.

Examples showcase various styles of anime character generation effects.

Scene Generation

Generates anime-style scenes based on text descriptions.

Examples showcase scenes such as Victorian mansions and carnivals.

Content Creation

Concept Design

Provides rapid concept design for anime, games, and other projects.

🚀 GGUF quantized and fp8 scaled version of hyvid with lora anime adapter

This project offers a GGUF quantized and fp8 scaled version of hyvid integrated with a lora anime adapter, enabling high - quality text - to - video generation.

🚀 Quick Start

📦 Installation

To set up the project, you need to drag the following files to their respective directories:

Drag hyvid_lora_adapter.safetensors [323MB] to ./ComfyUI/models/loras.
Drag hunyuan - video - t2v - 720p - q4_0.gguf [[7.74GB](https://huggingface.co/calcuis/hyvid/blob/main/hunyuan - video - t2v - 720p - q4_0.gguf)] to ./ComfyUI/models/diffusion_models.
Drag llava_llama3 - q4_0.gguf [[4.68GB](https://huggingface.co/calcuis/hyvid/blob/main/llava_llama3 - q4_0.gguf)] to ./ComfyUI/models/text_encoders.
Drag clip_l_fp8_e4m3fn.safetensors [123MB] to ./ComfyUI/models/text_encoders.
Drag hunyuan_video_vae_fp8_e4m3fn.safetensors [247MB] to ./ComfyUI/models/vae.

💻 Usage Examples

🏃‍♂️ Run it straight (no installation needed way)

Run the .bat file in the main directory (assuming you are using the gguf - node pack below).
Drag the demo clip or the workflow json file (below) to your browser.

📋 Workflows

Example workflow for [gguf](https://huggingface.co/calcuis/hyvid/blob/main/workflow - hyvid - gguf.json) (upgrade your node for llava gguf support; more quantized llava - llama3 version [here](https://huggingface.co/chatpig/llava - llama3/tree/main)).
Example workflow for [safetensors](https://huggingface.co/calcuis/hyvid/blob/main/workflow - hyvid - safetensors.json) (fp8 scaled version [13.2GB] is recommended).

🔍 Review

More stable output if adapter applied.
Working good even with fp8_e4m3fn scaled clip and vae.
Significant changes in loading speed while using the new quantized/scaled file(s) with revised workflow (see above).

📚 References

Lora adapter from [trojblue](https://huggingface.co/trojblue/HunyuanVideo - lora - AnimeShots).
Base model from tencent.
Fast model from fastvideo.
Comfyui from comfyanonymous.
Comfyui - gguf [city96](https://github.com/city96/ComfyUI - GGUF).
Gguf - comfy [pack](https://github.com/calcuis/gguf - comfy/releases).
Gguf - node ([pypi](https://pypi.org/project/gguf - node)|repo|pack).

📎 Appendices

Get more test prompt from the training [dataset](https://huggingface.co/datasets/calcuis/anime - descriptor).
Still no ideas? Get fresh prompt from our simulator.

📄 License

This project is under the MIT license named tencent - hunyuan - community.

screenshot

Property	Details
Model Type	GGUF quantized and fp8 scaled version of hyvid with lora anime adapter
Training Data	- trojblue/test - HunyuanVideo - anime - images - calcuis/anime - descriptor
Pipeline Tag	text - to - video
Tags	hyvid, lora, gguf - comfy, gguf - node

Featured Recommended AI Models

Empowering the Future, Your AI Solution Knowledge Base

English 简体中文繁體中文にほんご