convert mp4 to gif on ubuntu
ffmpeg -i formula-ocr-step3.mp4 -r 15 -vf "scale=680:-1" formula-ocr-step3.gif
convert images to webp:
for file in *.png *.jpeg .jpg; do cwebp "$file" -o "${file%.}.webp"; done for file in .mp4;do ffmpeg -i "$file" -c:v libvpx-vp9 -crf 32 -b:v 0 -cpu-used 4 -c:a libopus "${file%.}.webm"; done
batch replace to .webp
/assert/models/(.*).(png|jpe?g|webp) -> https://cdn.z-image.app/models/$1.webp
用户痛点收集
-
自动化转换与重定向(Retargeting)
- 骨骼绑定(Rigging) 预设骨骼支持:直接支持导出适配 Unity Humanoid、Unreal Engine 5 Manny/Quinn、Mixamo 的格式
- 重定向(Retargeting):,让生成的动作能自动适配到用户上传的常见模型(如 Mixamo 骨骼、UE5 Mannequin、VRM 模型)上
-
非人类模型 human motion
-
转换为视频
-
连贯性问题:AI 生成的动作往往是片段式的。如果用户需要“先走路,然后停下,再挥手”,目前的模型可能很难一次性生成流畅的长序列。你可能需要开发“动作融合”功能
-
Planning
- multiple language support
-
Implementation
- Create
components/models-uidirectory - Implement
ModelSelector
- Create
-
Verification
- Verify UI responsiveness and aesthetics
- Verify logic (Model inputs, validation, API integration)
-
UI Improvements
- playground history
- canvas Loop / Play / Pause 控制按钮
-
Function complete(功能完备性)
- contact page submit
-
Performance: https://pagespeed.web.dev/analysis/https-z-image-app/7gz67k5eoh?form_factor=mobile
- improve performance
- prompts
- gallery
- https://z-image.pro/pricing 功能最完善 模型支持多
- https://z-image.co/#ai-image-generator 多模型支持
- https://z-image.ai/ 嵌入hf 链接
- https://zimage.top/pricing 嵌入hf链接
- https://zimage.net/ 嵌入hf链接
domain: z-image.app
model information:
- huggingface: https://huggingface.co/Tongyi-MAI/Z-Image-Turbo
- official hompage: https://tongyi-mai.github.io/Z-Image-homepage/
wsl 开发
pnpm next dev -H 0.0.0.0
prisma
pnpm add -D prisma pnpm add @prisma/client pnpm prisma generate # 生成 prisma client
prisma change schema
npx prisma migrate dev --name 描述这次改动的名字
vercel deploy build config
pnpm prisma generate && pnpm build
google auth: https://console.cloud.google.com/auth/clients/create?hl=zh-cn&project=gen-lang-client-0746857282
pnpm add next@15.3.6 react@19.1.2 react-dom@19.1.2
- favicon 保留svg格式
- cludflare 关闭 ai crawl control