Voice Message Text Copy: Getting the Words Out

2026-10-25

Short answerCopy it or download the TXT. Those are the two ways out, and they hand you the same string. We ran a 20-second voice message through six different files today and got 307 characters, 62 words every single time, with 0 newline characters in the result. The SRT and VTT buttons add 33 and 39 bytes to that same text and no timing information at all, so TXT is the one to use unless a player file is what you need.

What you are actually copying

One line. Today's 20-second message came back as 307 characters and 62 words with 0 newline characters, 0 timestamps, 0 speaker labels and no confidence score. Here is the whole thing:

"Hey, it's me. I just wanted to say I'm running about 10 minutes late. Traffic on the bridge is completely stopped. I'll text you again when I'm close. Don't order yet. I want to look at the menu first. Also, can you bring the charger? Mine is at home. Thanks. See you in a bit. Oh, and if the place is full."

The model did three things you would have done by hand: it wrote "10" instead of "ten", expanded "it's" and "I'm", and put a period at the end of each sentence. It did not add a single line break. Paste that into a notes app and it lands as one paragraph.

Length: what 5 seconds gives you against 90

We cut one 29-second voice message into a ladder and uploaded each rung as a 16 kHz mono M4A at 64 kbps.

5 s, 44,947 B: 84 characters, 17 words, 1.23 s
10 s, 91,093 B: 162 characters, 30 words, 1.61 s
15 s, 135,532 B: 237 characters, 46 words, 1.88 s
20 s, 179,542 B: 307 characters, 62 words, 2.50 s
30 s, 266,742 B: 468 characters, 93 words, 4.55 s
60 s, 531,505 B: 937 characters, 187 words, 6.02 s
90 s, 796,898 B: 949 characters, 189 words, 6.86 s

Read that as 15 to 17 characters per second of audio. It holds across the ladder: 84 characters over 5 seconds is 16.8, 468 over 30 is 15.6. The 60- and 90-second rungs are the same 29-second message on a loop, so their word counts are a shape, not a benchmark.

Six files, one text

A voice message reaches you in whatever wrapper the app used, and people re-encode it before they send it anywhere. We took the same 20 seconds and wrote it out six ways.

M4A, 64 kbps, 179,542 B: 307 characters, 62 words, 2.50 s
MP3, 32 kbps, 80,819 B: 307 characters, 62 words, 3.40 s
OGG/Opus, 24 kbps, 55,629 B: 307 characters, 62 words, 3.41 s
WAV, 641,102 B: 307 characters, 62 words, 2.85 s
M4A, 128 kbps, 186,964 B: 307 characters, 62 words, 2.74 s
MP4 with video, 213,542 B: 307 characters, 62 words, 3.46 s

That is a 12-fold spread in file size - 55,629 bytes against 641,102 - and the text did not budge by one character. Which container you have is not worth thinking about. Which extension you have is: the endpoint checks the filename, so a file with no extension, or a .3gp or .amr one, gets refused before anything is read.

Character counts for one voice message cut to 5, 10, 15, 20 and 30 seconds, and the same 20 seconds rebuilt in six containers

What the three download buttons write

We ran the page's own download code in a real browser against today's 307-character text and measured what each button builds.

TXT: 307 bytes, 1 line
SRT: 340 bytes, 4 lines
VTT: 346 bytes, 5 lines

The SRT and VTT files carry the full text inside one cue that runs from 00:00:00,000 to 00:00:00,000 - a zero-length cue. Our response has no timing field, so there is nothing to fill those numbers with, and the page does not invent any. You gain 33 and 39 bytes of wrapper and no timing at all.

Getting it out

Copy the text if you are pasting it somewhere. Download the TXT if you want a file. Both give you the same 307 characters. Take SRT or VTT only when a video editor or a player wants that container - and expect to retime every line yourself.

If you want paragraphs, split on the periods after you paste. The transcript has none of its own, and a 60-second message comes back as a single 937-character block.

What we did not measure today

We measured the string the three buttons build, not the clipboard itself. The clips are synthetic voice recordings, not messages pulled out of the app, and real ones arrive with room tone, compression artefacts and the occasional notification sound on top.

Next step

Got a Reel you want as text? Upload the file and download the transcript as TXT, SRT, or VTT →

Affiliate link

Container was never the problem here: six files with a 12-fold spread in size all returned the same 307 characters. What Wondershare UniConverter fixes is the case where the message you have sits in a wrapper this tool refuses by name, or is too big to send — it re-exports to MP3 or M4A on your own device first.

FAQ

How do I copy the text of a voice message?

Upload the audio file, then use the copy button for the text itself or the TXT button for a file. Both hand you the same string. Today's 20-second message was 307 characters and 62 words, with 0 newline characters, so expect one paragraph.

Why is the transcript all one line?

The response carries the text and nothing else - no timestamp, segment or word list. Today's 307-character result had 0 newline characters in it. The TXT download is 307 bytes in a single line for exactly that reason.

Should I download SRT or VTT instead of TXT?

Only if a player or editor demands that container. Measured in a browser today, the same 307-character text came out as 307 bytes of TXT, 340 of SRT and 346 of VTT, and the SRT and VTT cues were 0.000 seconds long. You pay 33 or 39 bytes for a header and get no timing.

Does the file format change the transcript?

Not in the run we did today. The same 20 seconds came back as 307 characters and 62 words in M4A, MP3, OGG/Opus, WAV and MP4, across a 12-fold spread in file size. The filename extension matters more than the contents - unknown extensions are refused on name alone.