跳到正文
cccskills
免费在 GitHub 公开

dreamifly-batch

批量调用 Dreamifly 生成并下载图片或视频。

Use when the user asks to batch-generate images, edit images with references, create text-to-video or image-to-video, queue prompts, validate or estimate generation cost, retry failed media jobs, or summarize Dreamifly outputs.

中文简介正在准备中。当前显示原文说明。

shaozheng0503/dreamifly-batch

实测评分 未实测

查看安装方法

包含的文件（64）

  • SKILL.md2.9 KB
  • .claude/skills/dreamifly-batch/examples/prompts.jsonl317 B
  • .claude/skills/dreamifly-batch/references/evals.md463 B
  • .claude/skills/dreamifly-batch/references/gotchas.md421 B
  • .claude/skills/dreamifly-batch/references/model-selection.md567 B
  • .claude/skills/dreamifly-batch/references/output-contract.md340 B
  • .claude/skills/dreamifly-batch/SKILL.md1.5 KB
  • .cursor/skills/dreamifly-batch/examples/prompts.jsonl317 B
  • .cursor/skills/dreamifly-batch/references/evals.md463 B
  • .cursor/skills/dreamifly-batch/references/gotchas.md421 B
  • .cursor/skills/dreamifly-batch/references/model-selection.md567 B
  • .cursor/skills/dreamifly-batch/references/output-contract.md340 B
  • .cursor/skills/dreamifly-batch/SKILL.md1.5 KB
  • .gitignore334 B
  • AGENTS.md4.4 KB
  • config/config.json396 B
  • config/cookie.txt.example701 B
  • docs/architecture.md1.8 KB
  • docs/flow-overview.png878.5 KB
  • docs/flow-social-card.png961.8 KB
  • docs/flow-video.png833.3 KB
  • docs/flow-xiaohei.png905.5 KB
  • docs/sample-i2i-winter.png3.2 MB
  • docs/sample-i2v-garden.gif3.8 MB
  • docs/sample-i2v-garden.mp43.7 MB
  • docs/sample-japanese-garden.png2.9 MB
  • docs/sample-t2v-city-sunset.gif2.9 MB
  • docs/sample-t2v-city-sunset.mp44.5 MB
  • docs/screenshot-models-1.png1.6 MB
  • docs/screenshot-models-2.png1.1 MB
  • docs/skill-cases/ai-comic/ai_comic_21819ba90b46fc14.png1.4 MB
  • docs/skill-cases/ai-comic/ai_comic_a88cefdc57ceb6ba.png1.2 MB
  • docs/skill-cases/ai-comic/ai_comic_eddd52baf16afe9b.png1.2 MB
  • docs/skill-cases/ai-comic/ai_comic_ffaba646652c4c20.png1.2 MB
  • docs/skill-cases/case_xiaohei_gpt.png1.0 MB
  • docs/skill-cases/case-animation-video.png885.3 KB
  • docs/skill-cases/case-slide-deck.png1.1 MB
  • docs/skill-cases/case-social-card.png1.5 MB
  • docs/skill-cases/ppt-generic/ppt-concept-boat.png821.8 KB
  • docs/skill-cases/ppt-generic/ppt-hero-ribbon.png912.5 KB
  • docs/skill-cases/ppt-generic/ppt-section-desk.png472.0 KB
  • docs/skill-ecosystem.md17.6 KB
  • docs/skill-thinking.md6.4 KB
  • docs/style-anime.png901.2 KB
  • docs/style-lego.png989.4 KB
  • docs/style-lineart.png1.1 MB
  • docs/style-oil.png1.3 MB
  • docs/style-pixel.png818.0 KB
  • docs/style-riso.png1.7 MB
  • docs/xiaohei-integration.md6.3 KB
  • dreamify.py57.3 KB
  • examples/prompts.jsonl317 B
  • images/.gitkeep0 B
  • install.sh1.8 KB
  • LICENSE1.0 KB
  • prompts.example.txt494 B
  • README_EN.md8.5 KB
  • README.md24.3 KB
  • references/evals.md842 B
  • references/gotchas.md1.0 KB
  • references/model-selection.md1.2 KB
  • references/output-contract.md963 B
  • run.sh239 B
  • tests/test_dreamify.py2.4 KB

SKILL.md（原文）

安装前可以查看交给 AI 代理的指令内容。

Dreamifly 批量生图 / 生视频

把用户的提示词队列转成 Dreamifly 图片/视频生成任务，下载媒体到 images/，并用 results.jsonl / failed.jsonl 给 Agent 稳定汇报和重试。

Trigger Boundary

Use this skill for:

  • 批量生图、图生图编辑、统一排队出图
  • 文生视频、图生视频、多参考图生视频
  • 校验 prompt 队列、估算积分、重试失败项、汇总生成结果

Do not use this skill for generic image prompt writing if the user does not want Dreamifly generation, local files, or queue execution.

Execution Protocol

Always run this sequence for real jobs:

  1. Choose the model and write one task per line to prompts.txt, or use --prompts jobs.jsonl.
  2. Run python3 dreamify.py --validate.
  3. Run python3 dreamify.py --estimate.
  4. Run python3 dreamify.py --dry-run.
  5. Run python3 dreamify.py --check.
  6. If any item is 5+ credits, video, or nano-banana-2, stop and ask for explicit user approval.
  7. Run ./run.sh or python3 dreamify.py -n N.
  8. Report with python3 dreamify.py --summary, then inspect results.jsonl and run.log if needed.

Cheap image models (Wai-*, Z-Image-Turbo, Qwen-Image-Edit) can run without extra approval after validation passes.

Queue Formats

Inline text:

a neon cat | model=Z-Image-Turbo | 16:9 | x2 | seed=123 | neg=blurry
edit this, add snow | model=Qwen-Image-Edit | img=ref.png
city timelapse | model=happyhorse-1.0 | secs=5 | res=720P

JSONL:

{"prompt":"a neon cat in rain","model":"Z-Image-Turbo","aspectRatio":"16:9","batch_size":1}
{"prompt":"edit this, add snow","model":"Qwen-Image-Edit","images":["ref.png"]}

Required Safety Rules

  • Never print, commit, copy, or summarize config/cookie.txt; only say whether a cookie was loaded and its length if needed.
  • Never run video, nano-banana-2, or any 5+ credit item without explicit approval.
  • Do not blindly retry 401/402; ask the user to update the cookie or credits first.
  • Video jobs are expensive and must not be auto-retried.

Progressive References

Read these only when needed:

  • Model routing and parameters: references/model-selection.md
  • Output files and reporting contract: references/output-contract.md
  • Failure lessons and hard boundaries: references/gotchas.md
  • Skill routing eval cases: references/evals.md
  • JSONL examples: examples/prompts.jsonl

评价

还没有评价。欢迎分享你的使用感受。

评论

无需登录即可发表。欢迎分享使用感受、提问，或说明在哪些环境下无法运行。

还没有评论。来写第一条吧。

提交后会立即公开。请勿填写邮箱、电话等个人信息。URL 最多 2 个。

0 / 1000 字

此表单受 reCAPTCHA 保护，适用 Google 的隐私权政策和服务条款。

同一仓库的技能

dreamifly-batch

免费

Batch-generate Dreamifly images/videos from prompt queues. Use when the user asks to batch-generate images, edit images with references, create text-to-video or image-to-video, validate or estimate Dreamifly job cost, retry failed media jobs, or summarize Dreamifly outputs.

中文简介正在准备中。当前显示原文说明。

shaozheng0503/dreamifly-batch22026年6月26日 更新

查看 shaozheng0503 的全部技能

报告此技能的问题