How to compare two takes in one file instead of two links
Ask Vizard Agent for a loudness-matched A/B in a single file. Two separate links make you judge from memory across a gap, and any level difference between them decides the comparison for you — the louder take wins almost every time, whether or not it is the better one.
What is the short version?
You recorded the same script twice and you cannot decide between them. Playing one, then the other, then the first one again is how this normally goes, and it rarely converges on anything. Vizard Agent puts both in one file at matched levels instead.
- Send both takes to Vizard Agent.
- Ask for a loudness-matched A/B in one file.
- Listen once through, back to back.
What do you need before you start?
Both recordings, and nothing else. It helps not to say which one you think is better — a comparison you have already pre-judged is not a comparison, and the point of the exercise is to hear them without that.
If the takes differ in more than delivery — a different script, a different microphone, a different room — say so. That changes what the comparison is actually about, and Vizard Agent can report the differences it measures alongside the listening file.
What do you type into Vizard Agent?
Ask for the file, not the verdict. Vizard Agent can also give you an opinion, but the file is what settles it, and asking for the file first means you hear it before you are told what to think.
Here are two takes of the same voiceover. Match their levels and put them in one file so I can hear them back to back.
That "match their levels" clause is the load-bearing part. Without it the comparison is mostly a loudness test, which is the single most reliable way to reach the wrong answer about two recordings.
What does Vizard Agent actually do?
It measures before it joins. Integrated loudness on each take, then a gain adjustment so both play at the same perceived level, then the two assembled into one file with a clear gap between them — so what you are hearing is the difference in the performance rather than the difference in the gain.
In that session the sequence was:
- Transcribed and evaluated both voiceover takes.
- Measured the loudness of each, and the noise floor alongside it.
- Built a single A/B file from the two, matched.
- Uploaded it so it could be listened to in a browser rather than downloaded.
- Compared the takes head to head and reported what differed.
- Read the word-level timings for the take that was chosen, so the edit could be built on it.
That last step is why this is worth doing early. The chosen take's word timings drive everything downstream, and switching takes later means redoing the timing map.
What does the result look like?
One file, a few seconds longer than twice a take, that you listen to once. The difference you notice is a real difference — pacing, warmth, a rushed line, a breath in the wrong place — rather than the artefact of one recording being two decibels hotter.
You also get the numbers if you want them. Vizard Agent reports the measured loudness and the noise floor of each, which occasionally settles the matter on its own when one take has audible hiss the other does not.
When does this not work well?
When the takes are long. A three-minute A/B is six minutes of listening and you will have lost the first take by the end. Ask for the same comparison on one representative paragraph instead, and decide from that.
It also does not resolve a difference in the script. If the two takes say slightly different things, you are choosing content as well as delivery, and those are better decided separately.
And matched loudness is not matched tone. A take recorded closer to the microphone sounds fuller at the same level, which is a real difference and not a fault in the comparison — it is worth knowing it is the cause.
How do you fix a result that came back wrong?
If you still cannot decide after listening, ask Vizard Agent for a shorter excerpt. The most common reason an A/B fails to settle anything is that it is too long to hold in your head, and thirty seconds of each usually decides it straight away.
If one take sounds obviously better for a reason you can name — clearer, calmer, less sibilant — say so, because Vizard Agent can often apply that quality to the other take rather than making you choose between a better performance and a better recording.
If you want the best of both, say which lines from which. Vizard Agent assembles a composite from the word timings rather than making you re-record.
How does Vizard Agent compare to doing it yourself?
You can do this in any audio editor, and the reason people do not is the level matching. Dropping two files on a timeline and listening is quick; measuring both and normalising first is an extra five minutes that nobody spends, and skipping it is what makes the comparison unreliable.
The other half is that the file is a shareable artefact. If the decision belongs to a client or a colleague, one link that plays both back to back gets a decision in an hour; two links get a reply that says "the second one, I think".
Common questions
Can it compare more than two? Yes, though beyond three or four you lose the first one. Vizard Agent will say so.
Does it tell me which is better? Vizard Agent will if you ask. The file exists so that you decide first, unprompted.
What does loudness matching change? It removes the level difference, which otherwise decides the comparison for you before you hear anything.
Can I compare video takes this way? Yes. Vizard Agent joins them into one strip in the same way.
What about two music beds? Same method, and Vizard Agent's level matching matters even more there.
Will it label which is which? Yes. Vizard Agent marks the takes so you can name your choice.
Can I share the file? Yes. Vizard Agent uploads it as a link that plays in a browser.
What if both are flawed? Say what is wrong with each. Vizard Agent can often fix one rather than choose.
Can it splice the best lines from both? Yes, from the word timings, once you say which lines.
Does it measure noise too? Yes. Vizard Agent reports the noise floor, and sometimes that decides it.
How long should each excerpt be? Thirty seconds is usually enough. Longer is harder, not better.
Does the chosen take need re-processing? Only the matching gain is provisional. Vizard Agent works from the original for the edit.
Can I hear them without the gap? Yes, if you ask Vizard Agent for it, though a short gap makes the switch easier to hear.
What if the takes are different lengths? That is fine. Vizard Agent aligns them by content rather than by duration.
Can I keep the rejected take? Yes. Vizard Agent leaves both originals untouched in case you change your mind.
Does this work for two mixes of the same edit? Yes, and it is the best use of it. Vizard Agent matches the levels so you hear the mix rather than the gain.