Edition #001 · October 2026 · free to cite
The AI Video Tools Scorecard
Every "best AI video tool" list is a shelf. This is a scorecard: each tool graded A–C per job, because a tool that aces clipping flunks editing. Sourcing is labeled on every row — we claim hands-on only what we ran ourselves.
| Tool | Job | Grade | One-line verdict | Sourced |
|---|---|---|---|---|
| Opus Clip | Long video → shorts drafts | A− | Best-in-class first drafts; cull its output, don't trust the virality score | hands-on |
| Opus Clip | Precision editing | C | Not an editor; fighting it for frame control wastes your afternoon | hands-on |
| Descript | Talking-head / podcast editing | A− | Transcript editing is a real 2–3× speedup; Studio Sound is the sleeper | hands-on |
| Descript | Clip discovery ("find my best bits") | C+ | Asking it to be Opus is asking the wrong product | hands-on |
| CapCut | Free editing + captions | A | The $0 default that three of the paid tools barely beat on polish | hands-on |
| Riverside | Remote recording quality | A− | Local recording per guest; wifi stops deciding your audio fate | docs+community |
| Klap | Auto-shorts (Opus-class) | B | Simpler, cheaper feel; thinner polish reports from creators | docs+community |
| Vidyo.ai | Clips + light scheduling | B− | Convenient bundle; scheduling half is basic | docs+community |
| Munch | Marketing-team clipping | B− | Trend-matching angle; priced past solo creators | docs+community |
| ElevenLabs | AI voice / overdub quality | A− | The voice benchmark others get compared to | docs+community |
Grades: A = buy-first confidence · B = solid for its lane · C = wrong tool for that job (may ace another). Edition #001 baseline; re-graded as we complete hands-on passes. Reddit community-sentiment edition in progress.
How to cite this
The AI Cut — "The AI Video Tools Scorecard," Edition #001, October 2026, theaicut.xyz/scorecard
Free to cite and screenshot with attribution. Numbers and grades change as tools change — link rather than republish, so readers see the current edition.
Methodology
Hands-on rows: tool run on the same 38-minute talking-head benchmark footage (October 2026). Docs+community rows: vendor documentation cross-checked against consistent creator reports — flagged, never disguised as testing. No vendor input, no sponsored placement. Disclosure.