| # Build In Public Drafts |
|
|
| The hackathon has an optional Build in Public challenge. Use these as draft posts and adjust the tone before publishing. |
|
|
| ## Post 1 - Project Start |
|
|
| ### X / Twitter |
|
|
| Building ElevenClip.AI for the AMD Developer Hackathon. |
|
|
| It turns long-form videos into short-form clips with a human-AI editing loop: |
|
|
| - Whisper Large V3 for transcription |
| - Qwen2.5 for highlight scoring |
| - Qwen2-VL for visual signals |
| - ROCm + AMD MI300X target deployment |
|
|
| Creators should not need to watch a 2-hour video just to find 10 good clips. |
|
|
| @lablab @AIatAMD |
|
|
| ### LinkedIn |
|
|
| I am building ElevenClip.AI for the AMD Developer Hackathon. |
|
|
| The project is an AI clip studio that helps creators turn long-form videos into short-form clips for TikTok, YouTube Shorts, and Instagram Reels. The core workflow combines Whisper transcription, Qwen highlight detection, optional Qwen2-VL visual understanding, ffmpeg rendering, and a human-in-the-loop editor. |
|
|
| The production target is AMD Developer Cloud with ROCm and AMD Instinct MI300X, because long-form video processing needs high-throughput model inference and fast rendering. |
|
|
| ## Post 2 - Technical Update |
|
|
| ### X / Twitter |
|
|
| Technical update on ElevenClip.AI: |
|
|
| The local MVP now has: |
|
|
| - FastAPI backend |
| - React clip editor |
| - Channel profile inputs |
| - Upload/YouTube pipeline |
| - Mock transcript/highlight path for demo mode |
| - Clip cards with trim, subtitle edit, regenerate, approve, download |
| - Hugging Face Space live under the hackathon org |
|
|
| Next: run the real Whisper + Qwen pipeline on AMD Developer Cloud with ROCm and benchmark CPU vs MI300X. |
|
|
| @lablab @AIatAMD |
|
|
| ### LinkedIn |
|
|
| Technical update for ElevenClip.AI: |
|
|
| The MVP now has a FastAPI backend, React editor, channel profile setup, upload/YouTube input, transcript and highlight output, clip generation, and a human review interface. I also published a Hugging Face Space under the AMD Developer Hackathon organization. |
|
|
| The next milestone is the real AMD cloud run: Whisper Large V3 on ROCm PyTorch, Qwen2.5 through a ROCm-compatible serving path, and benchmark logs comparing CPU and AMD Instinct MI300X performance. |
|
|
| ## AMD Feedback Notes |
|
|
| Fill this after using AMD Developer Cloud: |
|
|
| - What was easy: |
| - What was confusing: |
| - ROCm setup notes: |
| - PyTorch/Transformers compatibility notes: |
| - vLLM ROCm notes: |
| - Benchmark result: |
| - Suggestion for AMD Developer Cloud documentation: |
|
|
|
|
