Agent skill

Sparkdiffusion Inference

by AlibabaResearch in AlibabaResearch/SparkDiffusion

Run validated SparkDiffusion inference for Wan 2.1 or Wan 2.2 T2V/I2V models.

Apache-2.0Auto-check passedMedia & Creative

Install Sparkdiffusion Inference

skills CLI

$ npx skills add AlibabaResearch/SparkDiffusion --skill sparkdiffusion-inference -a claude-code

Project install by default; add -g for ~/.claude/skills/.

GitHub CLI

$ gh skill install AlibabaResearch/SparkDiffusion sparkdiffusion-inference --agent claude-code

Project scope by default; add --scope user for a personal install. Needs GitHub CLI 2.90.0 or later (public preview).

Manual copy

$ git clone --depth 1 https://github.com/AlibabaResearch/SparkDiffusion.git skills-src && mkdir -p .claude/skills && cp -r skills-src/.agents/skills/sparkdiffusion-inference .claude/skills/sparkdiffusion-inference && rm -rf skills-src

Use ~/.claude/skills/ instead of .claude/skills for a personal install. The folder must contain SKILL.md.

Claude Code skills documentation · loads skills from .claude/skills/

Paste into Claude Code

Install the "sparkdiffusion-inference" agent skill from https://github.com/AlibabaResearch/SparkDiffusion/tree/main/.agents/skills/sparkdiffusion-inference into .claude/skills/sparkdiffusion-inference/ in this project. Copy the whole folder (SKILL.md and every file beside it), keep the folder name "sparkdiffusion-inference", then confirm the skill loads.

Claude Code copies the folder itself, the same result as the manual copy. Check what it changed before you commit it.

$skill-installer

$skill-installer install https://github.com/AlibabaResearch/SparkDiffusion/tree/main/.agents/skills/sparkdiffusion-inference

Type this inside Codex. $skill-installer <name> installs a curated skill from openai/skills. The installer writes to $CODEX_HOME/skills (default ~/.codex/skills). Restart Codex if the skill does not show up.

skills CLI

$ npx skills add AlibabaResearch/SparkDiffusion --skill sparkdiffusion-inference -a codex

Project install goes to .agents/skills/; add -g for ~/.codex/skills/.

GitHub CLI

$ gh skill install AlibabaResearch/SparkDiffusion sparkdiffusion-inference --agent codex

Project scope by default (.agents/skills/); add --scope user for a personal install.

Manual copy

$ git clone --depth 1 https://github.com/AlibabaResearch/SparkDiffusion.git skills-src && mkdir -p .agents/skills && cp -r skills-src/.agents/skills/sparkdiffusion-inference .agents/skills/sparkdiffusion-inference && rm -rf skills-src

Use ~/.agents/skills/ instead of .agents/skills for a personal install.

Codex skills documentation · loads skills from .agents/skills/

Paste into Codex

Install the "sparkdiffusion-inference" agent skill from https://github.com/AlibabaResearch/SparkDiffusion/tree/main/.agents/skills/sparkdiffusion-inference into .agents/skills/sparkdiffusion-inference/ in this project. Copy the whole folder (SKILL.md and every file beside it), keep the folder name "sparkdiffusion-inference", then confirm the skill loads.

Codex copies the folder itself, the same result as the manual copy. Check what it changed before you commit it.

skills CLI

$ npx skills add AlibabaResearch/SparkDiffusion --skill sparkdiffusion-inference -a cursor

Project install goes to .agents/skills/; add -g for ~/.cursor/skills/.

GitHub CLI

$ gh skill install AlibabaResearch/SparkDiffusion sparkdiffusion-inference --agent cursor

Project scope by default (.agents/skills/); add --scope user for a personal install.

Manual copy

$ git clone --depth 1 https://github.com/AlibabaResearch/SparkDiffusion.git skills-src && mkdir -p .cursor/skills && cp -r skills-src/.agents/skills/sparkdiffusion-inference .cursor/skills/sparkdiffusion-inference && rm -rf skills-src

Use ~/.cursor/skills/ instead of .cursor/skills for a personal install.

Cursor skills documentation · loads skills from .cursor/skills/, .agents/skills/, .claude/skills/, .codex/skills/

Paste into Cursor

Install the "sparkdiffusion-inference" agent skill from https://github.com/AlibabaResearch/SparkDiffusion/tree/main/.agents/skills/sparkdiffusion-inference into .cursor/skills/sparkdiffusion-inference/ in this project. Copy the whole folder (SKILL.md and every file beside it), keep the folder name "sparkdiffusion-inference", then confirm the skill loads.

Cursor copies the folder itself, the same result as the manual copy. Check what it changed before you commit it.

gemini skills install

$ gemini skills install https://github.com/AlibabaResearch/SparkDiffusion.git --path .agents/skills/sparkdiffusion-inference

--scope user (default) or --scope workspace; --path is the subfolder of the repo that holds the skill; --consent skips the security confirmation prompt.

skills CLI

$ npx skills add AlibabaResearch/SparkDiffusion --skill sparkdiffusion-inference -a gemini-cli

Project install goes to .agents/skills/; add -g for ~/.gemini/skills/.

GitHub CLI

$ gh skill install AlibabaResearch/SparkDiffusion sparkdiffusion-inference --agent gemini-cli

Project scope by default (.agents/skills/); add --scope user for a personal install.

Manual copy

$ git clone --depth 1 https://github.com/AlibabaResearch/SparkDiffusion.git skills-src && mkdir -p .gemini/skills && cp -r skills-src/.agents/skills/sparkdiffusion-inference .gemini/skills/sparkdiffusion-inference && rm -rf skills-src

Use ~/.gemini/skills/ instead of .gemini/skills for a personal install, then run /skills reload.

Gemini CLI skills documentation · loads skills from .gemini/skills/, .agents/skills/

Paste into Gemini CLI

Install the "sparkdiffusion-inference" agent skill from https://github.com/AlibabaResearch/SparkDiffusion/tree/main/.agents/skills/sparkdiffusion-inference into .gemini/skills/sparkdiffusion-inference/ in this project. Copy the whole folder (SKILL.md and every file beside it), keep the folder name "sparkdiffusion-inference", then confirm the skill loads.

Gemini CLI copies the folder itself, the same result as the manual copy. Check what it changed before you commit it.

GitHub CLI

$ gh skill install AlibabaResearch/SparkDiffusion sparkdiffusion-inference

Installs for Copilot at project scope by default; add --scope user for a personal install. Preview a skill first with gh skill preview. Needs GitHub CLI 2.90.0 or later (public preview).

skills CLI

$ npx skills add AlibabaResearch/SparkDiffusion --skill sparkdiffusion-inference -a github-copilot

Project install goes to .agents/skills/; add -g for ~/.copilot/skills/.

Manual copy

$ git clone --depth 1 https://github.com/AlibabaResearch/SparkDiffusion.git skills-src && mkdir -p .github/skills && cp -r skills-src/.agents/skills/sparkdiffusion-inference .github/skills/sparkdiffusion-inference && rm -rf skills-src

Use ~/.copilot/skills/ instead of .github/skills for a personal install. Commit .github/skills so cloud agent and code review can use it.

GitHub Copilot skills documentation · loads skills from .github/skills/, .claude/skills/, .agents/skills/

Paste into GitHub Copilot

Install the "sparkdiffusion-inference" agent skill from https://github.com/AlibabaResearch/SparkDiffusion/tree/main/.agents/skills/sparkdiffusion-inference into .github/skills/sparkdiffusion-inference/ in this project. Copy the whole folder (SKILL.md and every file beside it), keep the folder name "sparkdiffusion-inference", then confirm the skill loads.

GitHub Copilot copies the folder itself, the same result as the manual copy. Check what it changed before you commit it.

skills CLI

$ npx skills add AlibabaResearch/SparkDiffusion --skill sparkdiffusion-inference -a opencode

OpenCode documents no install command of its own. Project install goes to .agents/skills/; add -g for ~/.config/opencode/skills/.

GitHub CLI

$ gh skill install AlibabaResearch/SparkDiffusion sparkdiffusion-inference --agent opencode

Project scope by default (.agents/skills/); add --scope user for a personal install.

Manual copy

$ git clone --depth 1 https://github.com/AlibabaResearch/SparkDiffusion.git skills-src && mkdir -p .opencode/skills && cp -r skills-src/.agents/skills/sparkdiffusion-inference .opencode/skills/sparkdiffusion-inference && rm -rf skills-src

Use ~/.config/opencode/skills/ instead of .opencode/skills for a personal install.

OpenCode skills documentation · loads skills from .opencode/skills/, .claude/skills/, .agents/skills/

Paste into OpenCode

Install the "sparkdiffusion-inference" agent skill from https://github.com/AlibabaResearch/SparkDiffusion/tree/main/.agents/skills/sparkdiffusion-inference into .opencode/skills/sparkdiffusion-inference/ in this project. Copy the whole folder (SKILL.md and every file beside it), keep the folder name "sparkdiffusion-inference", then confirm the skill loads.

OpenCode copies the folder itself, the same result as the manual copy. Check what it changed before you commit it.

Facts

Skill name: sparkdiffusion-inference
GitHub stars: 507
Token cost: ~853 tokens
SKILL.md length: 284 words
Files: 2
Skills in repo: 4
Repo updated: Oct 5, 2026
First seen: Oct 7, 2026
Licence: Apache-2.0

At a glance

Run validated SparkDiffusion inference for Wan 2.1 or Wan 2.2 T2V/I2V models.

A user asks to generate videos
SKILL.md covers Choose the Correct Path, Wan 2.1 T2V, Wan 2.1 I2V and Wan 2.2 T2V, plus 1 more section
Calls bash
Compare dense and sparse checkpoints

What it does

Sparkdiffusion Inference is an agent skill from AlibabaResearch/SparkDiffusion. Run validated SparkDiffusion inference for Wan 2.1 or Wan 2.2 T2V/I2V models. Use when a user asks to generate videos, compare dense and sparse checkpoints, choose distilled versus diffusion sampling, or debug checkpoint/path errors.

Its SKILL.md is about 850 tokens, which your agent loads only when the skill is triggered. The skill folder holds 2 other files (for example `agents/openai.yaml`).

It sits in Media & Creative, covering AI video generation. The licence is Apache-2.0.

When your agent uses it

A user asks to generate videos
Compare dense and sparse checkpoints
Choose distilled versus diffusion sampling
Debug checkpoint/path errors

Example prompts

“/sparkdiffusion-inference”

Requirements

Python 3

What it can do on your machine

Read from SKILL.md and the folder at commit caad1df. It shows what the files ask for, not the result of running them.

Tool permissions
Pre-approves nothing: there is no allowed-tools line, so your agent's usual permission prompts apply.
From allowed-tools in the SKILL.md frontmatter.
Runs code
Shell commands in SKILL.md call:
- bash
From the folder's file list and the shell code blocks in SKILL.md.
Network
No URLs in SKILL.md.
From URLs in SKILL.md, links to its own repository left out.
Credentials
Names no API keys, tokens, secrets or passwords.
From names ending in _API_KEY, _TOKEN, _SECRET, _KEY or _PASSWORD in SKILL.md.

Context cost

Sparkdiffusion Inference loads about 853 tokens when it runs. Until then it costs about 65 tokens; SKILL.md has 284 words of instructions outside code blocks.

Always · name and description, kept in context so the agent knows when to use it: ~65
When it runs · the whole SKILL.md, loaded when a task matches: ~853

Estimates: characters ÷ 4, the usual rule of thumb; real counts depend on the model's tokenizer. Scripts and assets cost tokens only if the agent reads them.

Safety

Auto-check passed

The automated check found no risky patterns in SKILL.md.

Automated static check — not a guarantee. Review scripts before installing. It scans the text of SKILL.md for risky patterns (piping downloads into a shell, reading credential files, hidden Unicode, destructive commands); files beside SKILL.md are not scanned.

SKILL.md

The full file from AlibabaResearch/SparkDiffusion at commit caad1df, republished under its Apache-2.0 licence (© AlibabaResearch). 284 words, ~853 tokens.

Download SKILL.mdSave it as .claude/skills/sparkdiffusion-inference/SKILL.md (or your agent's skills folder). This skill also uses 1 other file; get the full folder from GitHub.

name: sparkdiffusion-inference
description: Run validated SparkDiffusion inference for Wan 2.1 or Wan 2.2 T2V/I2V models. Use when a user asks to generate videos, compare dense and sparse checkpoints, choose distilled versus diffusion sampling, or debug checkpoint/path errors.

SparkDiffusion Inference

Use the public shell wrappers under scripts/inference/. They resolve model assets, validate paths, choose the correct Python entrypoint, and write outputs under a user-selected directory.

Choose the Correct Path

Use *_distilled.sh for a few-step distilled student.
Use *_diffusion.sh for the original multi-step teacher/base model.
Use the 2pt1 wrappers for Wan 2.1.
Use the 2pt2 wrappers for Wan 2.2.
Supplying an image to the Wan 2.1 wrapper switches it to I2V.
Wan 2.2 uses separate high- and low-noise experts; set CKPT_LOW when they are stored separately.

Do not feed a distilled checkpoint to a diffusion wrapper. The result can be noise even when the checkpoint and code are valid.

Standard dense and fused RoLa inference use repository-provided kernels and do not require SLA_SRC. Set it only for PureSLA, explicitly unfused RoLa, or legacy INT8 sparse-attention paths.

Wan 2.1 T2V

Distilled student:

bash

CUDA_VISIBLE_DEVICES=0 \
bash scripts/inference/eval_student_2pt1_distilled.sh \
  pretrain_weights/Wan2.1-T2V-14B-Diffusers \
  outputs/inference/wan21_t2v \
  4 fp8 14B_rola "" "" \
  "A cat playing in the garden under the sun."

Dense or sparse diffusion baseline:

bash

CUDA_VISIBLE_DEVICES=0 \
bash scripts/inference/eval_student_2pt1_diffusion.sh \
  pretrain_weights/Wan2.1-T2V-14B-Diffusers \
  outputs/inference/wan21_diffusion \
  50 bf16 14B "" "" \
  "A cat walking through a sunlit garden."

The topk argument is a keep ratio: 0.1, 0.05, and 0.03 mean 90%, 95%, and 97% sparsity respectively.

Wan 2.1 I2V

Pass the reference image as the seventh positional argument:

bash

bash scripts/inference/eval_student_2pt1_distilled.sh \
  pretrain_weights/Wan2.1-I2V-14B-480P-Diffusers \
  outputs/inference/wan21_i2v \
  4 fp8 14B_rola "" examples/i2v_input_1.jpg \
  "A person walks through a forest."

The checkpoint root must contain the matching VAE, text encoder, tokenizer, and image encoder, unless overridden with VAE_PATH, TEXT_ENCODER, TOKENIZER, or CLIP_ENCODER.

Wan 2.2 T2V

Distilled path:

bash

CKPT_LOW=pretrain_weights/Wan2.2-T2V-A14B-Diffusers/transformer_2 \
bash scripts/inference/eval_student_2pt2_distilled.sh \
  pretrain_weights/Wan2.2-T2V-A14B-Diffusers \
  outputs/inference/wan22_t2v \
  4 fp8 A14B_rola "" \
  "A cat playing in the garden under the sun."

The total step count is split between the high- and low-noise experts. On memory-constrained GPUs, follow the wrapper's expert paging behavior rather than loading both large experts permanently.

Preflight and Debugging

Before launching:

bash

source scripts/env.sh
bash -n scripts/inference/*.sh
test -e pretrain_weights/<model-root>

If the resolver cannot find an asset, either fix the repository layout or set the documented override variable. Do not change source code for a local path problem.

Use NUM_FRAMES, RESOLUTION, ASPECT_RATIO, SEED, NUM_SAMPLES, OUT_ROOT, and FIXED_RESOLUTION as environment overrides. Keep generated videos outside git-tracked source files.

Files

SKILL.md and 1 other file in .agents/skills/sparkdiffusion-inference of AlibabaResearch/SparkDiffusion.

SKILL.md
agents/openai.yaml

Open the folder on GitHubat commit caad1df

Compare with similar skills

Sparkdiffusion Inference next to the 5 skills that share the most tags, products or categories with it. Stars are the repository's; “used in” counts other GitHub owners with a copy.

Sparkdiffusion Inference compared with similar skills
Skill	Stars	Used in	Tokens	Auto-check	Licence	Repo updated
Sparkdiffusion Inference this skillAlibabaResearch/SparkDiffusion	507	—	~853	Automated safety check: Pass	Apache-2.0	3 days ago
Video Generationbytedance/deer-flow	83k	4 repos	~1.4k	Automated safety check: Pass	MIT	today
Video Shotseternityspring/reelbench-skills	868	2 repos	~1.8k	Automated safety check: Notes	Apache-2.0	17 days ago
Video Cover Imageitwanger/toBeBetterJavaer	18k	—	~3.3k	Automated safety check: Pass	None	today
Seedancesongguoxs/seedance-prompt-skill	2.9k	1 repos	~2.5k	Automated safety check: Pass	None	7 mo ago
HyperFrames Video Entry Pointheygen-com/hyperframes	59k	3 repos	~5.2k	Automated safety check: Pass	Apache-2.0	today

Similar skills

Video Generation
bytedance/deer-flow
Generates short videos from a structured JSON prompt, optionally guided by a reference image used as the first or last frame.
83k GitHub starsUsed in 4 repos~1.4k tokens
Media & CreativeAuto-check passed
Video Shots
eternityspring/reelbench-skills
拉片：把一条成片拆成逐镜头的分析表——每个镜头的时长、景别、类别、运镜、画面. An agent skill from eternityspring/reelbench-skills.
868 GitHub starsUsed in 2 repos~1.8k tokens
Media & CreativeAuto-check: notes
Video Cover Image
itwanger/toBeBetterJavaer
Generate matched 3:4, 16:9, and 4:3 short-video cover images from toBeBetterJavaer video scripts or AI/Java technical topics.
18k GitHub stars~3.3k tokensUpdated today
Media & CreativeAuto-check passed
Seedance
songguoxs/seedance-prompt-skill
This skill should be used when the user asks to "generate video prompts", "create Seedance prompts", "write video descriptions", mentions "Seedance", "seedance", "即梦", "即梦平台", "视频提示词", "视频生成"…
2.9k GitHub starsUsed in 1 repo~2.5k tokens
Media & CreativeAuto-check passed
HyperFrames Video Entry Point
heygen-com/hyperframes
Entry point for making, editing and rendering videos from HTML compositions with HyperFrames, routing each request to the right workflow.
59k GitHub starsUsed in 3 repos~5.2k tokens
Media & CreativeAuto-check passed
Lanshu Create AI Presenter Video
cclank/lanshu-create-ai-presenter-video
Turn a topic or finished script into a complete, publish-ready explainer video — led by an AI presenter from an authorized adult presenter image, or performed in one of nine visual explainer styles…
2.5k GitHub stars~3.6k tokensUpdated 2 days ago
Media & CreativeAuto-check passed

Questions about Sparkdiffusion Inference

What does Sparkdiffusion Inference do?

Run validated SparkDiffusion inference for Wan 2.1 or Wan 2.2 T2V/I2V models. Sparkdiffusion Inference is an agent skill from AlibabaResearch/SparkDiffusion.2 T2V/I2V models.

When should I use Sparkdiffusion Inference?

Sparkdiffusion Inference fits situations like: A user asks to generate videos; compare dense and sparse checkpoints; choose distilled versus diffusion sampling; debug checkpoint/path errors.

How do I install Sparkdiffusion Inference in Claude Code?

Run `npx skills add AlibabaResearch/SparkDiffusion --skill sparkdiffusion-inference -a claude-code`. Or copy the skill folder (.agents/skills/sparkdiffusion-inference in AlibabaResearch/SparkDiffusion) into .claude/skills/sparkdiffusion-inference in your project. Claude Code loads it when a task matches its description.

How do I install Sparkdiffusion Inference in Codex?

Run `npx skills add AlibabaResearch/SparkDiffusion --skill sparkdiffusion-inference -a codex`. Or copy the skill folder (.agents/skills/sparkdiffusion-inference in AlibabaResearch/SparkDiffusion) into .agents/skills/sparkdiffusion-inference in your project. Codex loads it when a task matches its description.

Can I use Sparkdiffusion Inference in Cursor, Gemini CLI or GitHub Copilot?

Cursor, Gemini CLI, GitHub Copilot and OpenCode also load SKILL.md folders. With the skills CLI, run `npx skills add AlibabaResearch/SparkDiffusion --skill sparkdiffusion-inference -a cursor` (or -a gemini-cli, github-copilot or opencode for the others). To copy it by hand, put the folder in .cursor/skills/sparkdiffusion-inference, .gemini/skills/sparkdiffusion-inference, .github/skills/sparkdiffusion-inference and .opencode/skills/sparkdiffusion-inference in your project.

What does Sparkdiffusion Inference need to run?

Going by SKILL.md and its folder, Sparkdiffusion Inference needs the command-line tools its instructions call (bash). Our summary lists: Python 3.

Does Sparkdiffusion Inference access the network?

SKILL.md contains no URLs. Any network use would come from the scripts or tools the agent runs. This is read from the text; nothing was executed.

Is Sparkdiffusion Inference safe to install?

Our automated static check of SKILL.md found no risky patterns, such as piping downloads into a shell, reading credential files or hidden Unicode. It is not a guarantee. Review the folder before installing.

What licence does Sparkdiffusion Inference use?

Sparkdiffusion Inference is published under the Apache-2.0 licence (the repository's licence). It allows redistribution, so the full SKILL.md is shown on this page.

How many tokens does Sparkdiffusion Inference use?

About 853 tokens (SKILL.md is roughly 3.4k characters). Agents keep only the skill's name and description in context until a task matches; then they load SKILL.md in full.

What are the alternatives to Sparkdiffusion Inference?

Skills that share tags, products or a category with Sparkdiffusion Inference: Video Generation (bytedance/deer-flow, 83k stars), Video Shots (eternityspring/reelbench-skills, 868 stars), Video Cover Image (itwanger/toBeBetterJavaer, 18k stars) and Seedance (songguoxs/seedance-prompt-skill, 2.9k stars). The comparison table on this page puts their stars, adoption, token cost, safety result and licence side by side.

Who maintains Sparkdiffusion Inference?

AlibabaResearch (a GitHub organization) maintains it in AlibabaResearch/SparkDiffusion, which has 507 GitHub stars. The repository holds 4 skills in this directory. The repository was last updated on October 5, 2026.

Source: AlibabaResearch/SparkDiffusion on GitHub. Facts on this page come from the repository at the commit we read; the author's words are quoted as theirs.

Sparkdiffusion Inference

Install Sparkdiffusion Inference

Facts

At a glance

What it does

When your agent uses it

Example prompts

Requirements

What it can do on your machine

Tool permissions

Runs code

Network

Credentials

Context cost

Safety

SKILL.md

SparkDiffusion Inference

Choose the Correct Path

Wan 2.1 T2V

Wan 2.1 I2V

Wan 2.2 T2V

Preflight and Debugging

Files

Compare with similar skills

Similar skills

Video Generation

Video Shots

Video Cover Image

Seedance

HyperFrames Video Entry Point

Lanshu Create AI Presenter Video

More from AlibabaResearch/SparkDiffusion

Sparkdiffusion Distill

Sparkdiffusion Finetune

Sparkdiffusion Setup

Questions about Sparkdiffusion Inference

Install Sparkdiffusion Inference

Facts

At a glance

What it does

When your agent uses it

Example prompts

Requirements

What it can do on your machine

Tool permissions

Runs code

Network

Credentials

Context cost

Safety

SKILL.md

SparkDiffusion Inference

Choose the Correct Path

Wan 2.1 T2V

Wan 2.1 I2V

Wan 2.2 T2V

Preflight and Debugging

Files

Compare with similar skills

Similar skills

More from AlibabaResearch/SparkDiffusion

Related

Questions about Sparkdiffusion Inference