Methodology
How We Rate AI Video Tools
The rubric, the sourcing rules, and the testing protocol we will hold ourselves to.
This site reviews AI video tools — their pricing, free tiers, safety posture and API access. Because no controlled hands-on test has been run, this page explains both how we verify facts today and how scores are produced. Nothing here is secret: every step is visible on the pages it affects.
The scoring rubric
Our rubric weights the things a buyer can verify and the things they cannot. These weights are fixed in advance and published here so a score cannot be tuned to reach a conclusion. Every overall score on this site is computed from these weights and the per-criterion scores in the table below — never set by hand.
| Criterion | Weight | What we measure |
|---|---|---|
| Price transparency | 25% | Published prices, published per-generation costs, no hidden expiry rules, clear refund terms |
| Model access | 20% | Breadth of the AI video/image model catalogue reachable on self-serve plans, including entry tiers |
| Output quality | 20% | Controlled hands-on testing uses the same fixed prompt set across tools, judged blind where possible; before then, clearly labelled editorial assessments of published showcase material may be used, with other untested products at neutral 5 |
| Free tier | 15% | What a non-paying user can actually accomplish: card requirements, quotas, watermarks, usable output |
| API access | 10% | Documented API or MCP access with published pricing, sane limits and self-serve keys |
| Billing clarity | 10% | Rollover rules, refund windows, renewal mechanics and cancellation friction, as written in the terms |
Weights are editorial policy, not sourced facts. They may change only with a visible note on this page.
How the overall is computed. Each criterion is scored 0-10. The overall score out of 10 is the weighted mean: overall = Σ(weight × score) ÷ 100, rounded to one decimal. Scoring scale: 0-2 poor, 3-4 weak, 5 neutral/untested, 6-7 fair, 8-9 strong, 10 exceptional.
| Criterion | AIContentDrop | Higgsfield | Artlist | InVideo |
|---|---|---|---|---|
| Price transparency (25%) | 8.0 | 7.0 | 6.0 | 7.0 |
| Model access (20%) | 8.0 | 8.0 | 8.0 | 7.0 |
| Output quality (20%) | 7.0 | 5.0 | 5.0 | 5.0 |
| Free tier (15%) | 5.0 | 5.0 | 6.0 | 6.0 |
| API access (10%) | 8.0 | 8.0 | 5.0 | 3.0 |
| Billing clarity (10%) | 7.0 | 6.0 | 6.0 | 6.0 |
| Overall | 7.3 | 6.5 | 6.1 | 6.0 |
AIContentDrop’s output-quality score of 7 is an editorial assessment of published showcase material, with controlled hands-on testing still pending; competitor output quality remains untested at the neutral 5. The overall is computed as Σ(weight × score) ÷ 100, never set by hand.
Scores are rubric-computed from published facts and labelled editorial assessments. AIContentDrop's output-quality score of 7 is an editorial assessment of its published showcase material; controlled hands-on testing is still pending and no hands-on test result is claimed. Competitor output-quality scores remain untested at the neutral 5.
How facts are verified today
- Every price, credit count, limit or policy claim must come from an official vendor page we fetched ourselves.
- Facts carry superscript citation markers; the numbered Sources section at the bottom of each page lists every URL with its checked date.
- Pricing figures for Higgsfield, Artlist and InVideo were re-read on the vendors' own pages on 6 October 2026 from a US vantage point; where a vendor geo-prices its page, we say so and show the other view too.
- Our screenshot captures of 30 September 2026 were taken from Cyprus and therefore show the EU visitor view; most display euros, while the InVideo capture displays its amounts with a $ sign. Each caption states which, so no capture is passed off as a current dollar table.
- If a number cannot be verified from an official page, we say it is not published — we do not guess, and we remove the claim rather than estimate it.
- Vendor promotions are described as time-boxed; promotional discounts and bonus credits are always called out as expiring.
- Worked cost examples are labelled illustrative math, never presented as vendor quotes.
The hands-on testing protocol (upcoming)
Once testing begins, each tool gets the same fixed prompt set — short social clips, product shots, human motion and text-heavy scenes — generated at the entry paid tier within one billing cycle. We record generation failures, effective clips per credit, watermark behaviour and export options. Test notes, dates and clip logs will be published alongside each score. Controlled testing will replace both AIContentDrop's current showcase-based editorial assessment and the competitors' untested neutral 5 output-quality scores with tested results. Until then, no output-quality score is represented as a controlled hands-on test result.
Editorial independence
Scores follow the rubric above and, once testing begins, the published protocol. Competitor facts are cited to the competitor’s own pages, and where one tool genuinely beats another on a criterion, the review says so. Corrections are welcome on the corrections page.
Update cadence
Pricing pages are re-checked monthly and matched to fresh captures; model and policy pages on change. Every page shows a visible last-verified date. When a cited number changes, the page is corrected and the change is logged on the corrections page.