🐕 Such Nostalgia. Very AI. Wow.

A Doge-Approved Time Machine — Powered by Alibaba Cloud AI

Much upload. Very rewind. Such video. Wow.

An AI-powered creative web app that transforms your modern photos into short nostalgic videos styled in the aesthetic of any internet era from 2010 to 2020. Pick a year, choose a festive greeting card theme, dial in an aesthetic mode — and let Qwen3.5 + Wan2.6-i2v-Flash do the rest.

Powered by Qwen3.5 and Wan2.6-i2v-Flash on Alibaba Cloud Model Studio, hosted on a Simple Application Server (SAS).

Built for the Alibaba Cloud AI × Creativity Hackathon 2026.

---

🐕 What It Does

1. Such Upload — drop any photo from your life in 2026
2. Very Pick — choose your era (2010–2020), festive theme (CNY / Ramadan / Eid), and aesthetic style (VSCO / Grunge / Meme / Cinemagraph)
3. Much Analyze — Qwen3.5 reads the photo and writes an era-appropriate scene description, vibe score, and Wan video direction prompt
4. Wow Generate — Wan2.6-i2v-Flash animates your photo with the chosen aesthetic
5. Such Result — you receive a short .mp4 + a Vibe Score (0–100) + the AI's full style prompt

---

🖥️ Demo

Live demo URL: https://your-sas-ip-or-domain

---

✨ Feature Highlights

📅 Era Selector

Choose which internet year to rewind to:

Era	Cultural Context
2010	Hipstamatic filters, early Instagram, flip-phone selfies, early Tumblr
2012	Earlybird/Hefe/X-Pro II, YOLO, Gangnam Style, Vine launch
2014	#nofilter, selfie sticks, Ice Bucket Challenge, early VSCO
2016 ★	VSCO cams, Doge meme, Harambe, Pokemon Go, dabbing (default)
2018	Avocado toast, VSCO girl, Drake memes, Instagram Stories
2020	Lo-fi aesthetic, quarantine lighting, TikTok grain, cottagecore

🎊 Festivity Selector (Greeting Card Mode)

Overlay a festive theme on top of the era aesthetic:

Theme	Vibe
✨ No theme	Pure internet nostalgia
🧧 Lunar New Year	Hongbao textures, firecracker bokeh, crimson-gold palette
🌙 Ramadan	Fanous lanterns, crescent moon overlays, amber-gold light
🎊 Eid	Arabesque patterns, fireworks bokeh, joyful golden glow

🎞️ Style Picker

Control the aesthetic mode that shapes the video generation:

Style	Description
Auto	Let Qwen3.5 decide based on the photo
VSCO	Warm fade, film grain, golden-hour light, #vscocam
Grunge	Low contrast, desaturated, dark moody texture
Meme-Heavy	Comic Sans overlays, reaction faces, maximum chaos
Cinemagraph	Film grain, anamorphic flare, letterbox feel

🤖 AI Transparency

The result page shows:

• Qwen's 2016 Reframe — the scene description the AI wrote
• Wan's Direction — the exact video generation prompt Qwen produced

---

🗂️ Project Structure

nostalgia-engine/
├── frontend/                  # Vue.js 3 single-page app
│   ├── src/
│   │   ├── components/        # UploadZone, StatusPoller, VideoResult
│   │   ├── views/             # Landing, Processing, Result
│   │   └── App.vue
│   └── Dockerfile
│
├── backend/                   # Python 3.11 + FastAPI
│   ├── main.py                # API routes
│   ├── models.py              # Pydantic request/response models
│   ├── job_store.py           # In-memory job store
│   ├── services/
│   │   ├── qwen.py            # Qwen3.5 dynamic prompt builder
│   │   ├── wan.py             # Wan2.6-i2v-Flash async job handling
│   │   └── oss.py             # OSS upload + signed URL generation
│   ├── requirements.txt
│   └── Dockerfile
│
├── infra/                     # OpenTofu infrastructure-as-code
│   ├── main.tf
│   ├── variables.tf
│   ├── outputs.tf
│   ├── terraform.tfvars.example
│   └── modules/
│       ├── sas/               # Simple Application Server instance + firewall
│       └── oss/               # Object Storage bucket + CORS + lifecycle
│
├── docs/
│   ├── ARCHITECTURE.md
│   └── screenshot.png
│
├── docker-compose.yml
├── .env.example
└── README.md

---

🧠 Architecture

User → Vue.js (SAS)
     → FastAPI Backend (SAS)
     → Qwen3.5            — analyzes photo + era/festivity/style → scene_description, style_prompt, vibe_score
     → Wan2.6-i2v-Flash   — generates nostalgic video (async, polled every 4s)
     → OSS                — stores .mp4, returns signed URL
     → User sees video + Vibe Score + AI prompt

See docs/ARCHITECTURE.md for the full sequence diagram and component breakdown.

---

🚀 Getting Started

Prerequisites

• Docker + Docker Compose
• Alibaba Cloud account with:
  • Simple Application Server (SAS) instance (Ubuntu 22.04, 2 vCPU / 4 GB RAM recommended)
  • Model Studio API key (access to qwen-vl-max and wan2.6-i2v-flash)
  • OSS bucket created

---

1. Clone the repository

git clone https://github.com/your-org/nostalgia-engine.git
cd nostalgia-engine

2. Configure environment variables

cp .env.example .env

Edit .env:

# Alibaba Cloud Model Studio
DASHSCOPE_API_KEY=sk-xxxxxxxxxxxxxxxxxxxxxxxx

# Models
QWEN_MODEL=qwen-vl-max
WAN_MODEL=wan2.6-i2v-flash

# OSS
OSS_ACCESS_KEY_ID=your_access_key_id
OSS_ACCESS_KEY_SECRET=your_access_key_secret
OSS_BUCKET_NAME=nostalgia-engine-outputs
OSS_ENDPOINT=oss-cn-hangzhou.aliyuncs.com
OSS_SIGNED_URL_EXPIRY=3600

3. Build and run

docker compose up -d --build

Service	Port
Frontend	http://localhost:80
Backend API	http://localhost:8000
API Docs (Swagger)	http://localhost:8000/docs

---

🔌 API Endpoints

POST /api/generate

Submit a photo for transformation.

curl -X POST http://localhost:8000/api/generate \
  -H "Content-Type: application/json" \
  -d '{
    "image_base64": "<base64>",
    "filename": "photo.jpg",
    "era": 2016,
    "festivity": "eid",
    "style": "vsco"
  }'

Field	Type	Required	Values
image_base64	string	✅	Base64-encoded JPEG/PNG/WebP
filename	string	✅	Original filename
era	integer	—	2010 2012 2014 2016 2018 2020 (default: 2016)
festivity	string	—	"cny" "ramadan" "eid" (default: null)
style	string	—	"vsco" "grunge" "meme" "cine" (default: null)

Response 202 Accepted:

{
  "job_id": "abc123",
  "status": "processing",
  "vibe_score": 78,
  "scene_description": "A casual outdoor moment remixed into 2016 VSCO aesthetic..."
}

---

GET /api/status/{job_id}

Poll for job completion.

curl http://localhost:8000/api/status/abc123

Response on success:

{
  "job_id": "abc123",
  "status": "success",
  "video_url": "https://your-bucket.oss-cn-hangzhou.aliyuncs.com/video/abc123.mp4?Expires=...",
  "vibe_score": 78,
  "scene_description": "A casual outdoor moment...",
  "style_prompt": "Slow warm zoom with golden bokeh drift, VSCO A4 fade..."
}

---

🛠️ Tech Stack

Layer	Technology
Frontend	Vue.js 3
Backend	Python 3.11 + FastAPI
Vision + Prompt AI	Qwen3.5 (Alibaba Cloud Model Studio)
Video Generation	Wan2.6-i2v-Flash (Alibaba Cloud Model Studio)
Storage	Alibaba Cloud OSS
Hosting	Alibaba Cloud SAS (Docker)
Infrastructure as Code	OpenTofu ≥ 1.6 + alicloud provider ~> 1.220
Dev Tooling	Qoder AI

---

⚙️ Configuration Reference

Variable	Description	Default
DASHSCOPE_API_KEY	Model Studio API key	—
QWEN_MODEL	Qwen model ID	qwen-vl-max
WAN_MODEL	Wan model ID	wan2.6-i2v-flash
OSS_BUCKET_NAME	OSS bucket for video output	—
OSS_ENDPOINT	OSS regional endpoint	oss-cn-hangzhou.aliyuncs.com
OSS_SIGNED_URL_EXPIRY	Signed URL validity in seconds	3600

---

🧪 Local Development (without Docker)

Backend:

cd backend
python -m venv venv && source venv/Scripts/activate
pip install -r requirements.txt
uvicorn main:app --reload --port 8000

Frontend:

cd frontend
npm install
npm run dev        # Vite dev server on :5173

---

📦 Deploying Infrastructure with OpenTofu

All cloud resources are declared in infra/ and provisioned with OpenTofu.

cd infra
cp terraform.tfvars.example terraform.tfvars
# fill in access keys, bucket name, SSH key pair
tofu init
tofu plan
tofu apply

OpenTofu provisions:

• OSS bucket — private, CORS enabled, 7-day video expiry lifecycle
• SAS instance — Ubuntu 22.04, Docker pre-installed, app auto-started via cloud-init
• Firewall rules — ports 22, 80, 8000

---

🪪 License

MIT — see LICENSE for details.

---

🙌 Acknowledgements

• Alibaba Cloud Model Studio — Qwen3.5 & Wan2.6-i2v-Flash
• Alibaba Cloud OSS
• Alibaba Cloud SAS
• Built with Qoder AI agentic coding tools

---

Alibaba Cloud AI × Creativity Hackathon 2026 — Such Nostalgia. Very AI. Wow. 🐕