Community resources
Follow each resource's requirements for its models, hardware, and xDiT version.
Projects and deployment guides
These resources have their own instructions and release versions.
| Resource | Source | Scope |
|---|---|---|
| xDiT ComfyUI | xdit-project | Runs supported diffusion models through xDiT's unified runner from ComfyUI. |
| Diffusion model inference | AMD-AGI | Provides benchmark configurations, profiling tools, tuning data, and container build infrastructure for diffusion model inference. |
| Serve Diffusion Transformer models using xDiT container on Cloud GPUs | Google Cloud | Documents xDiT container deployment in Model Garden, with FLUX.1-schnell examples. |
| xDiT diffusion inference | AMD ROCm AI Ecosystem | Documents container setup and diffusion model benchmarks on AMD Instinct gfx942 and gfx950 GPUs. |
For xDiT setup, see installation and containers. For workload configurations adapted to this CLI, see hardware recipes.
Model projects
StepVideo and Mochi use external model integrations rather than the unified runner. Follow each project's setup and version requirements.
| Project | Source | Scope |
|---|---|---|
| Step-Video-T2V | stepfun-ai | Provides StepVideo weights and inference code; its multi-GPU launcher uses xDiT's transformer TP implementation. |
| mochi-xdit: Parallel Inference for Mochi-preview Video Generation Model with xDiT | xdit-project | Provides a separate Mochi-preview inference implementation with Ulysses, Ring, and CFG parallelism. |
| SkyReels V1: Human-Centric Video Foundation Model | SkyworkAI | Provides SkyReels-V1 weights and its own SkyReelsInfer framework. For xDiT's SkyReels-V2 adapters, see the runner model list. |
Articles and case studies
Articles appear by publication date, newest first. Results apply to each article's hardware, software, and workload.
| Published | Article | Publisher | xDiT coverage |
|---|---|---|---|
| 2026-11-02 | How to Scale Diffusion Inference Across AMD RDNA 4 GPUs | ROCm Blogs | Shows xDiT image and video generation across one, two, or four Radeon AI PRO R9700S GPUs. |
| 2026-09-17 | Technical Dive into AMD MLPerf Inference v6.1 Submission | ROCm Blogs | Describes xDiT optimizations for Wan2.2-T2V-A14B within AMD's MLPerf Inference v6.1 submission. |