How We Test Tools
Last updated: 2026-05-02
Whenever we publish a comparison or "best" list, we've actually used the tools — recently, on real machines, against the same source material. This page documents exactly how, so you can reproduce our scoring and judge whether our picks would suit your use case.
Test rig
- Hardware: 2024 M3 MacBook Pro (16 GB) for desktop tests; iPhone 15 Pro and Pixel 8 for mobile tests. All with up-to-date OS and browsers as of the article's "last updated" date.
- Browsers: Chrome (latest), Safari (latest), Firefox (latest). Each test repeated in a fresh incognito session with a clean profile.
- Network: 100/100 Mbps wired connection. We re-run any borderline timing test on a throttled (10 Mbps down) profile to confirm the tool isn't bottlenecking on our network.
- Isolation: Browser-based tools are visited in a dedicated VM with a fresh profile so injected ads/tracking can't carry between tests.
Reference source material
To make tests comparable, we use the same Creative Commons-licensed source clips for every tool:
- Audio reference: a 3-minute CC BY 4.0 music track with known dynamic range and spectral content.
- Video reference: a 60-second CC BY 4.0 1080p clip containing motion, fine detail, and on-screen text (so encoder weaknesses become visible).
We never use copyrighted material we don't own for benchmarking, and we don't publish any output produced from copyrighted sources.
What we measure
| Dimension | How we score it |
|---|---|
| Speed | End-to-end wall-clock from URL paste to download-link ready, averaged over 3 runs at the same time of day. |
| Audio quality (MP3) | Spectrogram of output vs. source. We flag low-pass filtering, joint-stereo artefacts, and bit-padding (where the tool claims a bitrate it isn't actually producing). |
| Video quality (MP4) | Frame-by-frame visual comparison of source vs. output at matching resolution. We note codec used (H.264 / H.265 / VP9), frame rate, and visible re-encoding artefacts. |
| Ad / pop-up behaviour | Number of new tabs/pop-ups opened, redirects to ad networks, and any auto-installing browser-prompt activity. Counted across the full conversion flow. |
| Malware risk | Output file scanned with VirusTotal at test time. Tool domain checked against Google Safe Browsing and current uBlock Origin filter lists. |
| Privacy | Whether the tool requires sign-in, what trackers fire, what the privacy policy actually says. |
| Reliability | Whether the same URL works on three repeats spread over 24 hours, to catch tools whose source-platform integration is constantly broken. |
Scoring
Each tool is scored 0–10 on each dimension, weighted: speed 15%, audio quality 25%, video quality 15%, ad behaviour 15%, malware risk 15%, privacy 10%, reliability 5%. The composite score determines ranking. Where two tools tie within 0.3 points we break the tie on ad behaviour (the dimension users complain about most).
What disqualifies a tool
- Any malware flag from VirusTotal at test time.
- Auto-installing or pretending to require browser extensions.
- Forced sign-in or payment for a feature claimed to be free.
- Falsely advertising bitrates / resolutions the tool can't actually deliver (verified by spectrogram or by inspecting the output container).
Disclosure
We do not accept payment for placement in any "best of" list. Some entries link to third-party tools via affiliate links. When they do, the page carries a clear notice and the rank is determined entirely by the scoring above. See our affiliate disclosure.