Hello good people of the ffmpeg-users list, Because i'm not allowed to say anything in the #ffmpeg irc channel. Weird "cannot send to nick/channel" error message appears and thought i'd try the mailing list. Is there anyone who has experience with the silenceremove filter? I'm trying to remove long silences from recordings i made on my phone. The silences are actually a constant noise because of the microphone gain is set to 5.0x. (The playback meter in Audacity says the noise is around -48 to 42 dB). It is around 9 hours in duration. 1,1 GB large, signed 16 bit little endian mono 16 kHz wave file. Audacity says it's 32 bit float wave file though. To make sure that something is removed (or almost everything) i (hopefully) set the right silenceremove arguments to -20 db. # ffmpeg -i '/home/tony/test/Recording_53.wav' -af silenceremove=1:0:-20dB recording53_silences_removed.wav # Removes almost nothing. File almost as large as original. # Runs at incredible speed with e in the speed readout. # Done is a few minutes. # ffmpeg -i '/home/tony/test/Recording_53.wav' -af silenceremove=0:0:0:0:0:-1:0:-20dB:0:0:1:0 recording53_silences_removed.wav # Runs at 4.5x speed (takes long, like 1:30 hours) # leaves file of 500 MB and 4:30 hours in duration # still has silences of same level of several seconds, sometimes a minute. #ffmpeg -i '/home/tony/test/Recording_53.wav' -af silenceremove=start_periods=1:start_duration=0:start_threshold=-20dB:start_silence=3:star$ # Done in 40 seconds # leaves an empty file Does anyone know the right set of arguments for the silenceremove filter? Thanks a bunch.
Oh with this one: # ffmpeg -i '/home/tony/test/Recording_53.wav' -af silenceremove=0:0:0:0:0:-1:0:-20dB:0:0:1:0 recording53_silences_removed.wav # Runs at 4.5x speed (takes long, like 1:30 hours) # leaves file of 500 MB and 4:30 hours in duration # still has silences of same level of several seconds, sometimes a minute. ... this remarkable message appears: [out_0_0 @ 0x2a46500] 100 buffers queued in out_0_0, something may be wrong. [out_0_0 @ 0x2a46500] 1000 buffers queued in out_0_0, something may be wrong. Maybe has to do with the slowness...?
This one was truncated so i post it again #ffmpeg -i '/home/tony/test/Recording_53.wav' -af silenceremove=start_periods=1:start_duration=0:start_threshold=-20dB:start_silence=3:start_mode=any:stop_periods=2:stop_duration=3:stop_threshold=-20dB:stop_silence=3:stop_mode=any:detection=peak:window=10 recording53_silences_removed.wav # Done in 40 seconds # leaves an empty file
Is there anyone who has experience with the silenceremove filter? I'm trying to remove long silences from recordings i made on my phone. The silences are actually a constant noise because of the microphone gain is set to 5.0x. (The playback meter in Audacity says the noise is around -48 to 42 dB). It is around 9 hours in duration. 1,1 GB large, signed 16 bit little endian mono 16 kHz wave file. Audacity says it's 32 bit float wave file though.
Hey there, As you’ve noticed you pretty much need to know the basic “profile,” I guess, of the silence in your audio for best results. You could use the show volume filter with ffplay, % ffplay -f lavfi -i “amovie=$INPUT,showvolume" To get a general idea of what the noise is like, and apply the options accordingly. e.g. Use the levels of the silence/noise sections for the level options, setting -20db would cut out a lot more than noise, I think; see if there are short periods of noise, other than background noise, but still noise, and see how long those are to set a value for the duration options; also see if using peak and a window of 10 seconds makes sense (It might not, even if you set a more sensible value for the threshold, basically I think that would make it treat anything as silence if in a 10 second window it rises above the threshold even once)
On 2/19/20, Ted Park <kumowoon1025@gmail.com> wrote:
Is there anyone who has experience with the silenceremove filter? I'm trying to remove long silences from recordings i made on my phone. The silences are actually a constant noise because of the microphone gain is set to 5.0x. (The playback meter in Audacity says the noise is around -48 to 42 dB). It is around 9 hours in duration. 1,1 GB large, signed 16 bit little endian mono 16 kHz wave file. Audacity says it's 32 bit float wave file though.
Hey there, As you’ve noticed you pretty much need to know the basic “profile,” I guess, of the silence in your audio for best results. You could use the show volume filter with ffplay, % ffplay -f lavfi -i “amovie=$INPUT,showvolume" To get a general idea of what the noise is like, and apply the options accordingly. e.g. Use the levels of the silence/noise sections for the level options, setting -20db would cut out a lot more than noise, I think; see if there are short periods of noise, other than background noise, but still noise, and see how long those are to set a value for the duration options; also see if using peak and a window of 10 seconds makes sense (It might not, even if you set a more sensible value for the threshold, basically I think that would make it treat anything as silence if in a 10 second window it rises above the threshold even once)
If audio was in any way processed prior to silence removal, like in this case with gain, you will certainly get poor results.
Hello, Thanks so much for your reply. Would the microphone pickup just as much sounds (even faint ones) if i set gain back to 1.0x? If audio was in any way processed prior to silence removal, like in this case with gain, you will certainly get poor results.
On 2/19/20, To ny <tonynederpel@hotmail.com> wrote:
Hello, Thanks so much for your reply.
Would the microphone pickup just as much sounds (even faint ones) if i set gain back to 1.0x?
How are you supposed to differ from faint ones and noise itself. Get better microphone...
If audio was in any way processed prior to silence removal, like in this case with gain, you will certainly get poor results. _______________________________________________ ffmpeg-user mailing list ffmpeg-user@ffmpeg.org https://ffmpeg.org/mailman/listinfo/ffmpeg-user
To unsubscribe, visit link above, or email ffmpeg-user-request@ffmpeg.org with subject "unsubscribe".
Hey there, As you’ve noticed you pretty much need to know the basic “profile,” I guess, of the silence in your audio for best results. You could use the show volume filter with ffplay,
If audio was in any way processed prior to silence removal, like in this case with gain, you will certainly get poor results.
Morning, I’m not sure if you intended to reply to me, but I didn’t mean to suggest doing that if I have.
How do you figure results are correct?
ffmpeg -i '/home/tony/test/Recording_53.wav' -af silenceremove=1:0:-20dB recording53_silences_removed.wav this one does'nt remove anything.
ffmpeg -i '/home/tony/test/Recording_53.wav' -af silenceremove=0:0:0:0:0:-1:0:-20dB:0:0:1:0 recording53_silences_removed.wav this one leaves silences of more than a minute. Even though the thresholds are set to remove everything below -20dB.
ffmpeg -i '/home/tony/test/Recording_53.wav' -af silenceremove=start_periods=1:start_duration=0:start_threshold=-20dB:start_silence=3:start_mode=any:stop_periods=2:stop_duration=3:stop_threshold=-20dB:stop_silence=3:stop_mode=any:detection=peak:window=10 recording53_silences_removed.wav and this leaves nothing even though there as sounds louder than -20dB.
If you could provide some sample recordings (maybe if the ones you are working with are too sensitive to share, record some samples with the same settings you used with similar noise) it would make it easier to reproduce the results you are getting. I can only guess with just the commands you used and a qualitative description of the results. From the first command and result, I think maybe the recording doesn’t have silence at the beginning? From the second, since you gave 0 for the start_threshold, that is what will be used when removing silence from the middle sections. It’s not quite sure how anything is being removed at first glance to me, except for at the end. 0 would be used when, for example, you inserted a “silence” gap or region in Audacity or some DAW and you want to remove it. For the third, I don’t know what is going on, I would have expected at least 6 seconds to be left after filtering. Some example files and the unabridged console output for your commands would be useful, perhaps after increasing -loglevel .
Thanks again for your reply. The first command i copied directly from a forum post. The recording on my phone doesn't have a noise or sound that exceeds -20dB until 10 minutes 47 seconds. The output file left by the first command has the first 2,5 seconds missing and starts directly with a sound. Where in the recording on my phone that sound starts at 2,5 seconds. The second command as it seems left an output file that sounds like it plays at twice the speed. Coincidentally it's about half the size and duration. But everything seems to be still there and nothing is removed. Unfortunately i can't send a more extensive logging because it seems when -loglevel or -v is used i get no output in console at all. ________________________________ Van: ffmpeg-user <ffmpeg-user-bounces@ffmpeg.org> namens Ted Park <kumowoon1025@gmail.com> Verzonden: woensdag 19 februari 2020 15:54 Aan: FFmpeg user questions <ffmpeg-user@ffmpeg.org> Onderwerp: Re: [FFmpeg-user] silenceremove use
Hey there, As you’ve noticed you pretty much need to know the basic “profile,” I guess, of the silence in your audio for best results. You could use the show volume filter with ffplay,
If audio was in any way processed prior to silence removal, like in this case with gain, you will certainly get poor results.
Morning, I’m not sure if you intended to reply to me, but I didn’t mean to suggest doing that if I have.
How do you figure results are correct?
ffmpeg -i '/home/tony/test/Recording_53.wav' -af silenceremove=1:0:-20dB recording53_silences_removed.wav this one does'nt remove anything.
ffmpeg -i '/home/tony/test/Recording_53.wav' -af silenceremove=0:0:0:0:0:-1:0:-20dB:0:0:1:0 recording53_silences_removed.wav this one leaves silences of more than a minute. Even though the thresholds are set to remove everything below -20dB.
ffmpeg -i '/home/tony/test/Recording_53.wav' -af silenceremove=start_periods=1:start_duration=0:start_threshold=-20dB:start_silence=3:start_mode=any:stop_periods=2:stop_duration=3:stop_threshold=-20dB:stop_silence=3:stop_mode=any:detection=peak:window=10 recording53_silences_removed.wav and this leaves nothing even though there as sounds louder than -20dB.
If you could provide some sample recordings (maybe if the ones you are working with are too sensitive to share, record some samples with the same settings you used with similar noise) it would make it easier to reproduce the results you are getting. I can only guess with just the commands you used and a qualitative description of the results.
From the first command and result, I think maybe the recording doesn’t have silence at the beginning?
From the second, since you gave 0 for the start_threshold, that is what will be used when removing silence from the middle sections. It’s not quite sure how anything is being removed at first glance to me, except for at the end. 0 would be used when, for example, you inserted a “silence” gap or region in Audacity or some DAW and you want to remove it.
For the third, I don’t know what is going on, I would have expected at least 6 seconds to be left after filtering. Some example files and the unabridged console output for your commands would be useful, perhaps after increasing -loglevel . _______________________________________________ ffmpeg-user mailing list ffmpeg-user@ffmpeg.org https://ffmpeg.org/mailman/listinfo/ffmpeg-user To unsubscribe, visit link above, or email ffmpeg-user-request@ffmpeg.org with subject "unsubscribe".
The attachment in your first reply went through fine, I can see it. But it is too simple, I mean I can remove silence in it pretty well with % ffplay Downloads/sample.wav -af "silenceremove=start_periods=1:start_threshold=-30dB" But I assume you’re having more trouble with the original file because there’s occasional noise in the “silence” you want to remove, and the silence isn’t just at the beginning. Something more representative of the file you’re working with would be more helpful.
The first command i copied directly from a forum post. The recording on my phone doesn't have a noise or sound that exceeds -20dB until 10 minutes 47 seconds.
The output file left by the first command has the first 2,5 seconds missing and starts directly with a sound. Where in the recording on my phone that sound starts at 2,5 seconds.
The second command as it seems left an output file that sounds like it plays at twice the speed. Coincidentally it's about half the size and duration. But everything seems to be still there and nothing is removed.
I was confused by this at first since it doesn’t seem consistent with what you described before but is this the result when the sample you sent us is used with the same commands?
Unfortunately i can't send a more extensive logging because it seems when -loglevel or -v is used i get no output in console at all.
No output at all is weird, but did you set -loglevel $LEVEL where LEVEL is quiet, panic, fatal, error, warning, info, verbose, debug, or trace?
The sample is the best example since the thumps are quite soft and just a little louder than the noise. If all the three thumps and the voice would be left in the output, that would be great. I'll run your command and see what comes out. Ideally, it would be, if the thumps and voice would have some space between them during playback (of the output file), like three seconds apart and little bit of silence at the beginning. At the end silence isn't necessary. Ffplay i don't have compiled but using the silenceremove arguments with ffmpeg removes everything up until right before the two thumps with threshold set at -38dB. The first thump is removed unfortunately. The silence between the two thumps and my voice unfortunately isn't removed. At -37dB only my voice is left and none of the thumps are unfortunately left in. Just for clarification... In the whole recording is no absolute silence. Throughout the recording is noise of around -46dB. This is what i call "silence". ________________________________ Van: ffmpeg-user <ffmpeg-user-bounces@ffmpeg.org> namens Ted Park <kumowoon1025@gmail.com> Verzonden: donderdag 20 februari 2020 08:44 Aan: FFmpeg user questions <ffmpeg-user@ffmpeg.org> Onderwerp: Re: [FFmpeg-user] silenceremove use The attachment in your first reply went through fine, I can see it. But it is too simple, I mean I can remove silence in it pretty well with % ffplay Downloads/sample.wav -af "silenceremove=start_periods=1:start_threshold=-30dB" But I assume you’re having more trouble with the original file because there’s occasional noise in the “silence” you want to remove, and the silence isn’t just at the beginning. Something more representative of the file you’re working with would be more helpful.
The first command i copied directly from a forum post. The recording on my phone doesn't have a noise or sound that exceeds -20dB until 10 minutes 47 seconds.
The output file left by the first command has the first 2,5 seconds missing and starts directly with a sound. Where in the recording on my phone that sound starts at 2,5 seconds.
The second command as it seems left an output file that sounds like it plays at twice the speed. Coincidentally it's about half the size and duration. But everything seems to be still there and nothing is removed.
I was confused by this at first since it doesn’t seem consistent with what you described before but is this the result when the sample you sent us is used with the same commands?
Unfortunately i can't send a more extensive logging because it seems when -loglevel or -v is used i get no output in console at all.
No output at all is weird, but did you set -loglevel $LEVEL where LEVEL is quiet, panic, fatal, error, warning, info, verbose, debug, or trace? _______________________________________________ ffmpeg-user mailing list ffmpeg-user@ffmpeg.org https://ffmpeg.org/mailman/listinfo/ffmpeg-user To unsubscribe, visit link above, or email ffmpeg-user-request@ffmpeg.org with subject "unsubscribe".
On 2/20/20, To ny <tonynederpel@hotmail.com> wrote:
The sample is the best example since the thumps are quite soft and just a little louder than the noise.
Use peak level detector instead of rms one. Also reduce window size too.
If all the three thumps and the voice would be left in the output, that would be great.
I'll run your command and see what comes out.
Ideally, it would be, if the thumps and voice would have some space between them during playback (of the output file), like three seconds apart and little bit of silence at the beginning. At the end silence isn't necessary.
Ffplay i don't have compiled but using the silenceremove arguments with ffmpeg removes everything up until right before the two thumps with threshold set at -38dB. The first thump is removed unfortunately. The silence between the two thumps and my voice unfortunately isn't removed.
At -37dB only my voice is left and none of the thumps are unfortunately left in.
Just for clarification... In the whole recording is no absolute silence. Throughout the recording is noise of around -46dB. This is what i call "silence".
________________________________ Van: ffmpeg-user <ffmpeg-user-bounces@ffmpeg.org> namens Ted Park <kumowoon1025@gmail.com> Verzonden: donderdag 20 februari 2020 08:44 Aan: FFmpeg user questions <ffmpeg-user@ffmpeg.org> Onderwerp: Re: [FFmpeg-user] silenceremove use
The attachment in your first reply went through fine, I can see it. But it is too simple, I mean I can remove silence in it pretty well with % ffplay Downloads/sample.wav -af "silenceremove=start_periods=1:start_threshold=-30dB" But I assume you’re having more trouble with the original file because there’s occasional noise in the “silence” you want to remove, and the silence isn’t just at the beginning. Something more representative of the file you’re working with would be more helpful.
The first command i copied directly from a forum post. The recording on my phone doesn't have a noise or sound that exceeds -20dB until 10 minutes 47 seconds.
The output file left by the first command has the first 2,5 seconds missing and starts directly with a sound. Where in the recording on my phone that sound starts at 2,5 seconds.
The second command as it seems left an output file that sounds like it plays at twice the speed. Coincidentally it's about half the size and duration. But everything seems to be still there and nothing is removed.
I was confused by this at first since it doesn’t seem consistent with what you described before but is this the result when the sample you sent us is used with the same commands?
Unfortunately i can't send a more extensive logging because it seems when -loglevel or -v is used i get no output in console at all.
No output at all is weird, but did you set -loglevel $LEVEL where LEVEL is quiet, panic, fatal, error, warning, info, verbose, debug, or trace? _______________________________________________ ffmpeg-user mailing list ffmpeg-user@ffmpeg.org https://ffmpeg.org/mailman/listinfo/ffmpeg-user
To unsubscribe, visit link above, or email ffmpeg-user-request@ffmpeg.org with subject "unsubscribe". _______________________________________________ ffmpeg-user mailing list ffmpeg-user@ffmpeg.org https://ffmpeg.org/mailman/listinfo/ffmpeg-user
To unsubscribe, visit link above, or email ffmpeg-user-request@ffmpeg.org with subject "unsubscribe".
Ideally, it would be, if the thumps and voice would have some space between them during playback (of the output file), like three seconds apart and little bit of silence at the beginning. At the end silence isn't necessary.
Ffplay i don't have compiled but using the silenceremove arguments with ffmpeg removes everything up until right before the two thumps with threshold set at -38dB. The first thump is removed unfortunately. The silence between the two thumps and my voice unfortunately isn't removed.
At -37dB only my voice is left and none of the thumps are unfortunately left in.
Hi, Oh, sorry I misunderstood, I thought you wanted all noise cut out. I don’t know if this will work better since you want to include signals that are not much stronger than the background noise level, but have you maybe tried filters that reduce noise, like maybe anlmdn, rather than one that trims time-wise? For keeping pauses between, I think passing in the number of seconds for start_silence will do that for silenceremove, and you can trim the beginning if needed. If you want silence rather than the existing background noise, you could use silencedetect and use the metadata to generate and insert silence between sounds. Regards, Ted Park
Made a wave file of the first (softest) thump with the -accurate_seek, -t and -ss and one of the "relative" silence parts and as Ted suggested checked with showvolume: ffplay -f lavfi -i “amovie=sample.wav,showvolume" This way it's especially insightful to see the volume level of very short sounds. thump -37.8dB silence -48dB Used this one in ffmpeg now: silenceremove=start_periods=1:start_duration=0:start_threshold=-37dB:detection=peak:window=0 the step from -38dB to -37dB already nibbles off the leading silence and 36ms of the first thump. Going up to a certain threshold removs the first thump but silences of seconds are still left in. As it seems with the silenceremove i can't differentiate between the relative silence and the louder thump. I discovered the "truncate silence" in audacity now. It gets it done and you can just type in how long you want to leave silence in between sounds and leaves everything above threshold alone. Thanks anyway to Ted and Paul. ________________________________ Van: ffmpeg-user <ffmpeg-user-bounces@ffmpeg.org> namens Ted Park <kumowoon1025@gmail.com> Verzonden: vrijdag 21 februari 2020 04:53 Aan: FFmpeg user questions <ffmpeg-user@ffmpeg.org> Onderwerp: Re: [FFmpeg-user] silenceremove use
Ideally, it would be, if the thumps and voice would have some space between them during playback (of the output file), like three seconds apart and little bit of silence at the beginning. At the end silence isn't necessary.
Ffplay i don't have compiled but using the silenceremove arguments with ffmpeg removes everything up until right before the two thumps with threshold set at -38dB. The first thump is removed unfortunately. The silence between the two thumps and my voice unfortunately isn't removed.
At -37dB only my voice is left and none of the thumps are unfortunately left in.
Hi, Oh, sorry I misunderstood, I thought you wanted all noise cut out. I don’t know if this will work better since you want to include signals that are not much stronger than the background noise level, but have you maybe tried filters that reduce noise, like maybe anlmdn, rather than one that trims time-wise? For keeping pauses between, I think passing in the number of seconds for start_silence will do that for silenceremove, and you can trim the beginning if needed. If you want silence rather than the existing background noise, you could use silencedetect and use the metadata to generate and insert silence between sounds. Regards, Ted Park _______________________________________________ ffmpeg-user mailing list ffmpeg-user@ffmpeg.org https://ffmpeg.org/mailman/listinfo/ffmpeg-user To unsubscribe, visit link above, or email ffmpeg-user-request@ffmpeg.org with subject "unsubscribe".
On 2/21/20, To ny <tonynederpel@hotmail.com> wrote:
Made a wave file of the first (softest) thump with the -accurate_seek, -t and -ss and one of the "relative" silence parts and as Ted suggested checked with showvolume:
ffplay -f lavfi -i “amovie=sample.wav,showvolume" This way it's especially insightful to see the volume level of very short sounds.
thump -37.8dB silence -48dB
Used this one in ffmpeg now: silenceremove=start_periods=1:start_duration=0:start_threshold=-37dB:detection=peak:window=0
the step from -38dB to -37dB already nibbles off the leading silence and 36ms of the first thump. Going up to a certain threshold removs the first thump but silences of seconds are still left in.
As it seems with the silenceremove i can't differentiate between the relative silence and the louder thump.
I discovered the "truncate silence" in audacity now. It gets it done and you can just type in how long you want to leave silence in between sounds and leaves everything above threshold alone.
Thanks anyway to Ted and Paul.
You still have not learned anything about using ffmpeg and filter. Not setting stop periods to -1 will remove silence only from start. Good luck with audacity - it is big mess.
________________________________ Van: ffmpeg-user <ffmpeg-user-bounces@ffmpeg.org> namens Ted Park <kumowoon1025@gmail.com> Verzonden: vrijdag 21 februari 2020 04:53 Aan: FFmpeg user questions <ffmpeg-user@ffmpeg.org> Onderwerp: Re: [FFmpeg-user] silenceremove use
Ideally, it would be, if the thumps and voice would have some space between them during playback (of the output file), like three seconds apart and little bit of silence at the beginning. At the end silence isn't necessary.
Ffplay i don't have compiled but using the silenceremove arguments with ffmpeg removes everything up until right before the two thumps with threshold set at -38dB. The first thump is removed unfortunately. The silence between the two thumps and my voice unfortunately isn't removed.
At -37dB only my voice is left and none of the thumps are unfortunately left in.
Hi,
Oh, sorry I misunderstood, I thought you wanted all noise cut out. I don’t know if this will work better since you want to include signals that are not much stronger than the background noise level, but have you maybe tried filters that reduce noise, like maybe anlmdn, rather than one that trims time-wise? For keeping pauses between, I think passing in the number of seconds for start_silence will do that for silenceremove, and you can trim the beginning if needed. If you want silence rather than the existing background noise, you could use silencedetect and use the metadata to generate and insert silence between sounds.
Regards, Ted Park _______________________________________________ ffmpeg-user mailing list ffmpeg-user@ffmpeg.org https://ffmpeg.org/mailman/listinfo/ffmpeg-user
To unsubscribe, visit link above, or email ffmpeg-user-request@ffmpeg.org with subject "unsubscribe". _______________________________________________ ffmpeg-user mailing list ffmpeg-user@ffmpeg.org https://ffmpeg.org/mailman/listinfo/ffmpeg-user
To unsubscribe, visit link above, or email ffmpeg-user-request@ffmpeg.org with subject "unsubscribe".
Hello, and thanks so much for your reply. I was actually only requesting help for the silenceremove arguments (or settings or however they're called) on the commandline. The -20 dB threshold will not be my final settings. It was just to make sure something is removed. I don't know if you only read my additions to the original question (be replying to my first original email) or did read my original mail. I tried a few variations of the silenceremove filter arguments but with unexpected results. Kind regards, Tony ________________________________ Van: ffmpeg-user <ffmpeg-user-bounces@ffmpeg.org> namens Ted Park <kumowoon1025@gmail.com> Verzonden: woensdag 19 februari 2020 10:11 Aan: FFmpeg user questions <ffmpeg-user@ffmpeg.org> Onderwerp: Re: [FFmpeg-user] silenceremove use
Is there anyone who has experience with the silenceremove filter? I'm trying to remove long silences from recordings i made on my phone. The silences are actually a constant noise because of the microphone gain is set to 5.0x. (The playback meter in Audacity says the noise is around -48 to 42 dB). It is around 9 hours in duration. 1,1 GB large, signed 16 bit little endian mono 16 kHz wave file. Audacity says it's 32 bit float wave file though.
Hey there, As you’ve noticed you pretty much need to know the basic “profile,” I guess, of the silence in your audio for best results. You could use the show volume filter with ffplay, % ffplay -f lavfi -i “amovie=$INPUT,showvolume" To get a general idea of what the noise is like, and apply the options accordingly. e.g. Use the levels of the silence/noise sections for the level options, setting -20db would cut out a lot more than noise, I think; see if there are short periods of noise, other than background noise, but still noise, and see how long those are to set a value for the duration options; also see if using peak and a window of 10 seconds makes sense (It might not, even if you set a more sensible value for the threshold, basically I think that would make it treat anything as silence if in a 10 second window it rises above the threshold even once) _______________________________________________ ffmpeg-user mailing list ffmpeg-user@ffmpeg.org https://ffmpeg.org/mailman/listinfo/ffmpeg-user To unsubscribe, visit link above, or email ffmpeg-user-request@ffmpeg.org with subject "unsubscribe".
On 2/19/20, To ny <tonynederpel@hotmail.com> wrote:
Hello, and thanks so much for your reply.
I was actually only requesting help for the silenceremove arguments (or settings or however they're called) on the commandline.
The -20 dB threshold will not be my final settings. It was just to make sure something is removed.
I don't know if you only read my additions to the original question (be replying to my first original email) or did read my original mail. I tried a few variations of the silenceremove filter arguments but with unexpected results.
Silenceremove results are correct and your file(s) are broken.
Kind regards,
Tony
________________________________ Van: ffmpeg-user <ffmpeg-user-bounces@ffmpeg.org> namens Ted Park <kumowoon1025@gmail.com> Verzonden: woensdag 19 februari 2020 10:11 Aan: FFmpeg user questions <ffmpeg-user@ffmpeg.org> Onderwerp: Re: [FFmpeg-user] silenceremove use
Is there anyone who has experience with the silenceremove filter? I'm trying to remove long silences from recordings i made on my phone. The silences are actually a constant noise because of the microphone gain is set to 5.0x. (The playback meter in Audacity says the noise is around -48 to 42 dB). It is around 9 hours in duration. 1,1 GB large, signed 16 bit little endian mono 16 kHz wave file. Audacity says it's 32 bit float wave file though.
Hey there, As you’ve noticed you pretty much need to know the basic “profile,” I guess, of the silence in your audio for best results. You could use the show volume filter with ffplay, % ffplay -f lavfi -i “amovie=$INPUT,showvolume" To get a general idea of what the noise is like, and apply the options accordingly. e.g. Use the levels of the silence/noise sections for the level options, setting -20db would cut out a lot more than noise, I think; see if there are short periods of noise, other than background noise, but still noise, and see how long those are to set a value for the duration options; also see if using peak and a window of 10 seconds makes sense (It might not, even if you set a more sensible value for the threshold, basically I think that would make it treat anything as silence if in a 10 second window it rises above the threshold even once)
_______________________________________________ ffmpeg-user mailing list ffmpeg-user@ffmpeg.org https://ffmpeg.org/mailman/listinfo/ffmpeg-user
To unsubscribe, visit link above, or email ffmpeg-user-request@ffmpeg.org with subject "unsubscribe". _______________________________________________ ffmpeg-user mailing list ffmpeg-user@ffmpeg.org https://ffmpeg.org/mailman/listinfo/ffmpeg-user
To unsubscribe, visit link above, or email ffmpeg-user-request@ffmpeg.org with subject "unsubscribe".
How do you figure results are correct? ffmpeg -i '/home/tony/test/Recording_53.wav' -af silenceremove=1:0:-20dB recording53_silences_removed.wav this one does'nt remove anything. ffmpeg -i '/home/tony/test/Recording_53.wav' -af silenceremove=0:0:0:0:0:-1:0:-20dB:0:0:1:0 recording53_silences_removed.wav this one leaves silences of more than a minute. Even though the thresholds are set to remove everything below -20dB. ffmpeg -i '/home/tony/test/Recording_53.wav' -af silenceremove=start_periods=1:start_duration=0:start_threshold=-20dB:start_silence=3:start_mode=any:stop_periods=2:stop_duration=3:stop_threshold=-20dB:stop_silence=3:stop_mode=any:detection=peak:window=10 recording53_silences_removed.wav and this leaves nothing even though there as sounds louder than -20dB. Van: ffmpeg-user <ffmpeg-user-bounces@ffmpeg.org> namens Paul B Mahol <onemda@gmail.com> Verzonden: woensdag 19 februari 2020 12:16 Aan: FFmpeg user questions <ffmpeg-user@ffmpeg.org> Onderwerp: Re: [FFmpeg-user] silenceremove use On 2/19/20, To ny <tonynederpel@hotmail.com> wrote:
Hello, and thanks so much for your reply.
I was actually only requesting help for the silenceremove arguments (or
settings or
however they're called) on the commandline.
The -20 dB threshold will not be my final settings. It was just to make sure
something is removed.
I don't know if you only read my additions to the original question (be
replying
to my first original email) or did read my original mail. I tried a few
variations
of the silenceremove filter arguments but with unexpected results.
Silenceremove results are correct and your file(s) are broken.
Kind regards,
Tony
________________________________
Van: ffmpeg-user <ffmpeg-user-bounces@ffmpeg.org> namens Ted Park
<kumowoon1025@gmail.com>
Verzonden: woensdag 19 februari 2020 10:11
Aan: FFmpeg user questions <ffmpeg-user@ffmpeg.org>
Onderwerp: Re: [FFmpeg-user] silenceremove use
Is there anyone who has experience with the silenceremove filter? I'm
trying to remove long silences from recordings i made on my phone. The
silences are actually a constant noise because of the microphone gain is
set to 5.0x. (The playback meter in Audacity says the noise is around -48
to 42 dB). It is around 9 hours in duration. 1,1 GB large, signed 16 bit
little endian mono 16 kHz wave file. Audacity says it's 32 bit float wave
file though.
Hey there,
As you’ve noticed you pretty much need to know the basic “profile,” I guess,
of the silence in your audio for best results. You could use the show volume
filter with ffplay,
% ffplay -f lavfi -i “amovie=$INPUT,showvolume"
To get a general idea of what the noise is like, and apply the options
accordingly.
e.g. Use the levels of the silence/noise sections for the level options,
setting -20db would cut out a lot more than noise, I think; see if there are
short periods of noise, other than background noise, but still noise, and
see how long those are to set a value for the duration options; also see if
using peak and a window of 10 seconds makes sense (It might not, even if you
set a more sensible value for the threshold, basically I think that would
make it treat anything as silence if in a 10 second window it rises above
the threshold even once)
_______________________________________________
ffmpeg-user mailing list
ffmpeg-user@ffmpeg.org
To unsubscribe, visit link above, or email
ffmpeg-user-request@ffmpeg.org with subject "unsubscribe".
_______________________________________________
ffmpeg-user mailing list
ffmpeg-user@ffmpeg.org
To unsubscribe, visit link above, or email
ffmpeg-user-request@ffmpeg.org with subject "unsubscribe".
_______________________________________________ ffmpeg-user mailing list ffmpeg-user@ffmpeg.org https://ffmpeg.org/mailman/listinfo/ffmpeg-user To unsubscribe, visit link above, or email ffmpeg-user-request@ffmpeg.org with subject "unsubscribe".
participants (3)
-
Paul B Mahol -
Ted Park -
To ny