vidbenchmark.com - sample video naming convention
==================================================

Storage
-------
All sample renders are uploaded to Cloudflare R2 under:

  /global-storage/samples/{yyyy-mm-dd}/

where {yyyy-mm-dd} is the date of the benchmark run that produced the clip.

File name pattern
-----------------
  {model-slug}-{test-category}-{tagline-keyword}-{duration}-{resolution}-{year}.mp4

  model-slug      lowercase model id as used in /public/data/models.json
  test-category   vlogger | product-hero | motion-stress | lipsync | scene-cut
  tagline-keyword benchmark-vidbenchmark-com (or another keyword phrase)
  duration        integer seconds + "s"  (5s, 10s)
  resolution      720p | 1080p | 2k | 4k
  year            4-digit year of the run

Examples:
  kling-3-omni-vlogger-benchmark-vidbenchmark-com-5s-1080p-2026.mp4
  hailuo-h3-product-hero-vidbenchmark-com-5s-2k-2026.mp4

Poster frames use the identical name with a .jpg extension.

Machine IDs in JSON
-------------------
Each entry in samplePrompts[] carries a stable machine id:

  smp_{yyyymmdd}_{modelslug}_{category}_{Ns}{res}

Example:
  smp_20260804_kling-3-omni-1080p_vlogger_5s1080p

Metadata requirements
---------------------
Every rendered <video> element on a model detail page must carry:
  - title="{Model Name} AI video sample: vlogger talking head, 5s 1080p, vidbenchmark benchmark test 2026"
  - an adjacent <figcaption> with the same descriptive text
  - a Schema.org VideoObject JSON-LD block (name, description, contentUrl,
    uploadDate, duration PT5S, thumbnailUrl, publisher: vidbenchmark)

Adding a new sample
-------------------
1. Upload the .mp4 and .jpg to R2 using the pattern above.
2. Add the entry to REAL_SAMPLES in scripts/generate-models.mjs.
3. Run: bun run scripts/generate-models.mjs && bun run scripts/generate-llms.mjs