01
Which recordings should you not bother with?
Ones with little speech, since ranking works from what was said. Ones with heavy crosstalk, where several people talk over each other for most of the runtime. Wide group shots with six or more people, where no crop can find a subject. And recordings where the value is visual rather than spoken, such as a silent screen demonstration.
02
Trying a short sample first
If you are unsure whether a recording is a good fit, cut a ten or fifteen minute section that represents it well and try that first. It gives you an honest preview of the results without committing the whole file, and it shows quickly whether the recording suits short clips at all.
Choose a section from the middle of the recording rather than the start. Openings are often slower and less typical than the main part of a conversation.
03
Improving the next recording
Most difficult recordings share a few habits that are easy to change next time: several people speaking at once, a noisy room, a camera placed far from the speakers, or long stretches without speech. Small adjustments, such as a microphone per person and asking guests to take turns, change the outcome a great deal.
Write down what made the difficult recording hard, so the fix becomes part of your setup checklist.
04
Other uses for a poor fit
A recording that does not suit automatic short clips may still be valuable. It can be published in full, summarised in a written post, or used as the basis for a new, shorter recording where you restate the best points to camera.
Recording a two-minute summary yourself is often faster than wrestling with an unsuitable source.
Written notes from difficult recordings can also become written posts or newsletters, where overlapping speech and background noise matter far less than in a clip.
05
Setting expectations honestly
Knowing which recordings are a bad fit saves money and frustration. It is better to skip a file than to spend credits on results you will not post. When in doubt, try the sample first and decide based on what you see.
Over time you will recognise good material before you even upload it.
06
Getting help deciding
If you have a large archive and are not sure where to start, sort recordings by how clear the speech is and how focused the conversation was. Start with the clearest, most focused ones. They give the best results and teach you what to expect from the rest.
Leave the difficult files for last, or set them aside entirely.
07
Know the bad cases first
- Check whether speech carries the value. If it does not, set timestamps yourself on a paid plan.
- Be realistic about group shots. One or two people on camera is where the clean results come from.
- Test the awkward case for free. One video costs nothing, which is cheaper than assuming.
Test your most awkward recording on the free video. It costs nothing to find out.
Questions people ask
- What about a mostly instrumental recording?
- Poor fit for automatic selection. Manual timestamps on a paid plan handle it better.
- Eight people in a gallery grid?
- Be realistic. You get a readable crop of the grid rather than something made for vertical.
- Heavy background noise?
- Workable, but read the caption text before publishing.
- A silent screen recording?
- Ranking has little to go on, so choose the segments yourself.
- Is a bad fit still worth trying once?
- The first video is free, so the cost of finding out is nothing but time.