gemini-tts · voice "Kore", mentor tone), an acoustic instrumental bed (music), burned captions (local ffmpeg), and 5 real licensed clips fetched by stock-footage-fetch.
Pay yourself first · Track every dollar · Kill high-interest debt · Invest early · Live below your means.
| Script + keywords | per-beat VO line + concrete search keywords (the keywords describe the footage, not the narration) |
| Footage | stock-footage-fetch ×5 — real Pexels clips, ranked for 9:16 + duration. ~1–2s each, $0. |
| Voiceover | gemini-tts (voice "Kore", warm trustworthy-mentor tone) — one read per beat; caption + clip trimmed to its length |
| Captions | burned on-screen (faceless shorts are watched muted) |
| Finish | ffmpeg concat → 1080×1920 → instrumental music bed ducked under the VO → ffprobe gate |
faceless-stock-short playbook's flagship run. The same recipe powers the never-fail fallback: when an AI i2v job hangs, the server substitutes a stock clip like these — labeled and reversible.