| Feature | Needs |
|---|---|
| Caption transcription (word-timed captions) | nothing — local Whisper, free |
| Silence cut, music ducking, loudness | nothing — the first two follow the transcript, loudness is measured at render time |
| Keyword highlight in captions | optional — AI pick with a key, built-in heuristic without |
| Titles, descriptions, first comments, hooks, endcard lines | an AI key |
| Clip analysis (the clip briefs the generators) | an AI key |
| Caption translation per brand language | an AI key |
| Auto-fit "AI" / "reframe" modes | an AI key |
| AI voiceover for hooks/endcards | a TTS key |
No key, no AI: the app is fully usable without any of it. Every AI feature fails soft — a missing key or a refused call logs a warning and the render carries on.