Repromo

You vibe coded the app. We make the demo.

Paste a website or GitHub URL. A multi-agent showrunner on Qwen Cloud (Alibaba Cloud DashScope) writes the pitch, plans the scenes, and films real demo clips with HappyHorse.

Live app · Public repo · Architecture · License: MIT

---

Hackathon submission

Field	Value
Track	Track 2: AI Showrunner (Global AI Hackathon with Qwen Cloud)
Code repository (public)	https://github.com/fozagtx/repromo
Live demo	https://repromo.vercel.app
Open source license	MIT (detectable on the GitHub About panel)
Alibaba Cloud proof (code)	src/lib/qwen/client.ts · src/lib/video/happyhorse.ts
Architecture	See diagram below and docs/ARCHITECTURE.md
Demo video (~3 min)	Add public YouTube / Vimeo / Facebook Video URL here before submit

Proof of Alibaba Cloud / Qwen Cloud usage

Backend calls Alibaba Cloud DashScope (Qwen Cloud) APIs live. There is no mock mode.

Service	Model / API	Proof in repo
Qwen chat (scout, script, storyboard)	qwen-plus via DashScope OpenAI-compatible endpoint	src/lib/qwen/client.ts → https://dashscope-intl.aliyuncs.com/compatible-mode/v1
HappyHorse text-to-video	happyhorse-1.1-t2v async video synthesis	src/lib/video/happyhorse.ts → https://dashscope-intl.aliyuncs.com/api/v1/services/aigc/video-generation/video-synthesis

Orchestration: src/lib/agents/showrunner-graph.ts (LangGraph.js).

---

Project description

Repromo is an AI showrunner for builders who shipped an app but still need a launch / demo video.

1. You paste a public website or GitHub repo.
2. Agents read what you built, write a short pitch, plan cinematic scenes, then film real clips.
3. You get a playable demo video plus the script, with pause / stop controls while it runs.

Built for people who vibe-coded a product and need something they can post the same day - not a slide deck of mockups.

Features and functionality

• Site or GitHub input - works with product homepages and public repositories
• Multi-agent LangGraph pipeline - parse → scout → script → storyboard → generate_shots → finalize
• Qwen-powered creative - positioning, narration, and shot prompts grounded in real project context
• HappyHorse video generation - live DashScope async text-to-video (typically 2 shots, ~3-8s each at 720P)
• Job progress UI - stage chips, progress bar, pause / resume / stop
• Durable jobs - Neon Postgres so polling works across serverless instances
• Real logos in the input - favicon for the domain you paste
• Fail-fast without mocks - missing DASHSCOPE_API_KEY errors clearly; no fake videos

---

Architecture diagram

Clear path from frontend → backend → Qwen Cloud (Alibaba DashScope) → video → database → UI:

flowchart TB
  subgraph Client["Frontend"]
    UI["Repromo Next.js UI<br/>repromo.vercel.app"]
  end

  subgraph Backend["Next.js API / LangGraph"]
    GEN["POST /api/generate"]
    JOBS["GET/PATCH /api/jobs/:id"]
    GRAPH["Showrunner graph<br/>scout → script → storyboard → film"]
  end

  subgraph Alibaba["Alibaba Cloud · Qwen Cloud · DashScope"]
    QWEN["Qwen Plus<br/>compatible-mode/v1"]
    HH["HappyHorse 1.1 T2V<br/>video-synthesis + task poll"]
  end

  subgraph Data["Data"]
    NEON["Neon Postgres<br/>job status + results"]
    SRC["Website HTML or GitHub API"]
  end

  UI -->|paste URL| GEN
  GEN --> GRAPH
  GRAPH --> SRC
  GRAPH --> QWEN
  GRAPH --> HH
  GRAPH --> NEON
  UI -->|poll / pause / stop| JOBS
  JOBS --> NEON
  HH -->|video URL| UI

Loading

More detail: docs/ARCHITECTURE.md

Pipeline

parse_source → scout → script → storyboard → generate_shots → finalize

---

Demo video

Upload a public ~3 minute walkthrough to YouTube, Vimeo, or Facebook Video, then put the link here and on Devpost:

Video URL: https://youtu.be/YOUR_VIDEO_ID

Suggested chapters for judges:

1. Problem + paste a site/repo
2. Agents running (scout → script → storyboard → film)
3. Pause / stop controls
4. Final clips + script
5. Point at Alibaba Cloud proof files in the repo

---

Quick start (reproduce locally)

git clone https://github.com/fozagtx/repromo.git
cd repromo
npm install
cp .env.example .env

Required env (see .env.example):

DASHSCOPE_API_KEY=sk-...          # from https://home.qwencloud.com/api-keys
DATABASE_URL=postgresql://...     # Neon (or any Postgres) for job store

Optional:

QWEN_MODEL=qwen-plus
DASHSCOPE_VIDEO_MODEL=happyhorse-1.1-t2v

npm run dev

Open http://localhost:3000, paste e.g. linear.app or github.com/vercel/next.js, hit Make video.

Video rendering is async and can take several minutes. Keep the tab open; use Pause / Stop if needed.

---

Stack

Layer	Tech
Frontend	Next.js 16 App Router, HeroUI, Tailwind, Iconify
Agents	LangGraph.js + LangChain
LLM	Qwen (qwen-plus) on Alibaba DashScope
Video	HappyHorse (happyhorse-1.1-t2v) on DashScope
Jobs DB	Neon Postgres
Hosting	Vercel (app) + Alibaba Cloud DashScope (AI backend)

---

API

# start
curl -X POST https://repromo.vercel.app/api/generate \
  -H 'Content-Type: application/json' \
  -d '{"url":"https://linear.app"}'

# poll
curl https://repromo.vercel.app/api/jobs/<jobId>

# pause | resume | stop
curl -X PATCH https://repromo.vercel.app/api/jobs/<jobId> \
  -H 'Content-Type: application/json' \
  -d '{"action":"pause"}'

---

Project layout

src/
  app/                 # Landing UI + API routes
  components/ui/       # Navbar, prompt, cards, footer, logo
  lib/
    agents/            # LangGraph showrunner
    qwen/              # Alibaba DashScope chat client  ← proof
    video/             # HappyHorse create + poll         ← proof
    source/            # Website + GitHub ingest
    jobs/              # Neon-backed job store
docs/ARCHITECTURE.md
LICENSE                # MIT

---

License

MIT - open source. Copyright (c) 2026 Repromo.

---

Track 2: AI Showrunner · Global AI Hackathon with Qwen Cloud · fozagtx/repromo