logo
Docs

OpenFork Desktop

A decentralized AI video creation platform with compute marketplace and GitHub-style collaboration.

Download OpenFork Desktop
View Source Code

Integrated Services

Automated workflow orchestration, local Docker management, and seamless P2P job routing.

WAN 2.2
8-24GB

High-quality video (Text/Image-to-Video)

Hunyuan 1.5
16-24GB

Tencent's SOTA FP8 video model

LTX-2.3
12-32GB

LTX-2.3 22B: native audio+video generation (24GB)

DreamID-Omni
24GB

DreamID-Omni FP8 talking-head video with identity and voice reference inputs

Z-Image Turbo
8GB

ControlNet & Flux Image Gen

Z-Image
8-24GB

High-quality Z-Image (Q4_K_M GGUF)

Anima
8-16GB

Anima text-to-image illustration model

FLUX Kontext
8-24GB

FLUX.1 Kontext [dev] GGUF Q4_K_M for 8GB VRAM; supports optional two-image composition by precomposing the source and identity reference

Qwen Edit
8-12GB

Qwen-Image-Edit-2511 instruction-based image editing

Qwen Turbo
8GB

Ultra-fast instruction-based editing with an optional second reference image (2 steps)

VibeVoice
8GB

Clone-ready text-to-speech

Chatterbox
8GB

High-quality voice cloning

Qwen3-TTS
8-16GB

Alibaba's multilingual TTS with 9 premium voices

DiffRhythm
8GB

AI-powered music composition

Stable Audio
8GB

Sound FX and text-to-audio

AudioX
16-24GB

AudioX text-to-audio and video-conditioned sound generation

Local LLM
4GB

Qwen3 4B for workflow planning, scripts, prompts, and dialogue

TurboDiffusion
8-12GB

Real-time prototyping video model

HeartMuLa
16-24GB

4-bit quantized music generation (~8GB VRAM)

ACE-Step 1.5
8-16GB

ACE-Step 1.5 XL music generation for 16GB VRAM

MMAudio V2A
8-16GB

Video-to-audio synthesis (MMAudio small_44k)

Speech Enhancement
2GB

High-efficiency speech restoration and enhancement

PRiSM Audio
8-16GB

High-fidelity video-to-audio generation (PRiSM)

daVinci MagiHuman
16-32GB

daVinci-MagiHuman through Wan2GP with Magi Human Distill SR1080 quanto int8 weights. Requires a start image.

SCAIL
16-24GB

Experimental 16GB SCAIL 14B through Wan2GP for pose-guided character animation at reduced resolution and duration.

Vista4D
24GB

Vista4D 384p49 through Wan2GP for source-video novel-view reshooting with predefined camera trajectories.

SparkVSR
24GB

State-of-the-art video super-resolution (24GB VRAM required)

ERNIE-Image
16-24GB

Baidu ERNIE-Image text-to-image (FP16)

NVIDIA GPU ONLY 8GB+ VRAM RECOMMENDED