| --- |
| title: Simam Neural Codec Lab |
| emoji: 🎞️ |
| colorFrom: gray |
| colorTo: indigo |
| sdk: gradio |
| sdk_version: 5.0.0 |
| app_file: app.py |
| pinned: false |
| --- |
| |
| # Simam Neural Codec Lab |
|
|
| An evidence-first video compression observatory for immersive media. Upload a short clip to inspect its source properties and compare practical AV1, H.264, and HEVC encodes when the host FFmpeg build supports them. |
|
|
| This is a measurement baseline, not a claim that a learned codec is universally better. The next research adapter is planned around open projects such as CompressAI and Microsoft's DCVC family. |
|
|
| ## What it measures |
|
|
| - encoded size and bitrate; |
| - encode time and approximate real-time factor; |
| - decoded resolution and duration; |
| - a transparent quality proxy (PSNR when available); |
| - compatibility notes for XR playback. |
|
|
| Keep clips short while exploring. Results depend on the host FFmpeg build and are not a substitute for a controlled VMAF/LPIPS benchmark. |
|
|
| `examples/sample_codec.mp4` is a small synthetic clip for a repeatable first run. |
|
|
