Home |
Search |
Today's Posts |
#1
|
|||
|
|||
Recording and editing speech by phrases
Hello group,
I've been recording spoken text as a hobby with simple softwa Goldwave. If you have ever tried this you'll know that even if you are well prepared, eventually you'll make a mistake and you have to re-record part of the text. Of course, the ability to do a punch-in with preroll would improve the workflow, but there will of course be the annoyance of finding the exact position for the punch in. So: is there a simple piece of software that would allow the user to rewind and fast forward by "sentences" or "phrases" instead of linear time and to make punch-ins with "the start of phrase X" as the punch-in point? I know that the American Printing House for Blind People sell a program which seems to be capable of something like this, but the demo somehow did not convince me. (see: http://www.aph.org/tech/sr_info.htm). Besides, the software I'm looking for could be extremely simple. I'm not sure if a waveform display would be necessary. Regards, Ville Koskinen |
#2
|
|||
|
|||
|
#3
|
|||
|
|||
"Carey Carlan" wrote ...
(Ville Koskinen) wrote : So: is there a simple piece of software that would allow the user to rewind and fast forward by "sentences" or "phrases" instead of linear time and to make punch-ins with "the start of phrase X" as the punch-in point? I don't know of such a program, but there is a simpler solution. When you make a mistake that will require correction, repeat the sentence containing the mistake and continue. Exactly. Ever so much faster. And some people put some kind of audible marker on the flubs so they can be easily identified in review. I've heard syndicated radio programs where they missed one. :-) |
#4
|
|||
|
|||
Not sure if goldwave will do it, but I have used in in Wavelab... if
you pause, it will put markers on the scale at the top of the waveform window., that way after yuo flub, hit pause, then go back into record and continue. Later you can easily find your flubs, they will be right before the marker. then again, you could take a few moments then and just edit them out as you go. ;-) Richard Crowley wrote: "Carey Carlan" wrote ... (Ville Koskinen) wrote : So: is there a simple piece of software that would allow the user to rewind and fast forward by "sentences" or "phrases" instead of linear time and to make punch-ins with "the start of phrase X" as the punch-in point? I don't know of such a program, but there is a simpler solution. When you make a mistake that will require correction, repeat the sentence containing the mistake and continue. Exactly. Ever so much faster. And some people put some kind of audible marker on the flubs so they can be easily identified in review. I've heard syndicated radio programs where they missed one. :-) |
#5
|
|||
|
|||
When you make a mistake that will require correction, repeat the sentence
containing the mistake and continue. When you finish the recording, delete the fragment containing the error. This is really the cleanest and simplest way to handle that. I do this all the time when I'm reading script. If I stumble or mispronounce, I go back to the start of the sentence or paragraph and do another take and keep going, marking the spot on my script so I can edit it later while I listen to the playback. No need to shuttle, NTTAWWT. |
#6
|
|||
|
|||
Carey Carlan wrote:
I don't know of such a program, but there is a simpler solution. When you make a mistake that will require correction, repeat the sentence containing the mistake and continue. When you finish the recording, delete the fragment containing the error. Goldwave will do that with ease. Sure, I've been doing just that. Yet, I don't think this is such a non-issue you seem to imply. Regards, Ville Koskinen |
#7
|
|||
|
|||
"Ville Koskinen" wrote in message
oups.com... Carey Carlan wrote: I don't know of such a program, but there is a simpler solution. When you make a mistake that will require correction, repeat the sentence containing the mistake and continue. When you finish the recording, delete the fragment containing the error. Goldwave will do that with ease. Sure, I've been doing just that. Yet, I don't think this is such a non-issue you seem to imply. Regards, Ville Koskinen If you invent a better way, let us know. Here's another thing to think about. Back in the old analogue tape days I narrated long-form scripts for a producer who preferred to stop the tape whenever a mistake was made and punch in a new start at that point. At the end of the session he had a tape requiring no editing. Unfortunately, I always felt that the performance suffered. Rather than concentrating on the story telling, I became part of the mechanical process of creating a finished tape. In addition, the recording session was almost always twice as long as it needed to be, time for which he paid me. I think he lost on both his time wasted and my unnecessary additional time charges. However, I doubt that AT&T noticed the line item on their balance sheet.;-) An earlier poster mentioned the best answer I've found. In Adobe Audition I just tap F8, whenever I make a mistake. That places a marker on the time-line. Another key-stroke combination moves the cursor from one marker to the next in an instant. A few seconds takes care of each edit for the most part done visually on the wave-form. Steve King |
#8
|
|||
|
|||
Steve King wrote:
If you invent a better way, let us know. Here's another thing to think about. Back in the old analogue tape days I narrated long-form scripts for a producer who preferred to stop the tape whenever a mistake was made and punch in a new start at that point. At the end of the session he had a tape requiring no editing. Unfortunately, I always felt that the performance suffered. Rather than concentrating on the story telling, I became part of the mechanical process of creating a finished tape. Good points. I'm doing both narration and editing so I'll need almost two hours to produce an hour of edited reading. I too have noticed that having to hassle with the computer affects performance. A better way? What if by pressing let's say left arrow during recording, the program would stop recording and instantly start playing from the beginning of the last phrase you've spoken (phrase detection shouldn't be a problem). Pressing left arrow twice would rewind two phrases and so on. By pressing 'R' the program would punch in over the phrase you've selected, after a sufficient pre-roll, of course. Regards, Ville Koskinen |
#9
|
|||
|
|||
"Ville Koskinen" wrote in message
ups.com... A better way? What if by pressing let's say left arrow during recording, the program would stop recording and instantly start playing from the beginning of the last phrase you've spoken (phrase detection shouldn't be a problem). Pressing left arrow twice would rewind two phrases and so on. By pressing 'R' the program would punch in over the phrase you've selected, after a sufficient pre-roll, of course. "Phrase Detection" isn't as easy as you think. I suppose if you spoke like a robot and always took exactly the same amount of time between sentences it could be. But as Steve King pointed out the things you are wanting to do will more likely give you a worse performance. Cutting off your nose to spite your face, as it where. |
#11
|
|||
|
|||
|
#12
|
|||
|
|||
|
#13
|
|||
|
|||
|
#14
|
|||
|
|||
Ricky Hunt wrote:
"Phrase Detection" isn't as easy as you think. I suppose if you spoke like a robot and always took exactly the same amount of time between sentences it could be. But as Steve King pointed out the things you are wanting to do will more likely give you a worse performance. Cutting off your nose to spite your face, as it where. To me it seems that you only need to detect pauses, navigate through the audible stuff between them and make punch-ins. Perhaps you could even place the punch-in points so that they would be located just after the last "phrase", that is at the beginning of "silence". Then the reader could make sure that the pause before the rerecorded section would be exactly as long as he/she wanted. I don't know how this would affect performance. It seems to me that concentrating properly before making the correction would help. Regards, Ville Koskinen |
#15
|
|||
|
|||
"Ville Koskinen" wrote in message
oups.com... To me it seems that you only need to detect pauses, navigate through the audible stuff between them and make punch-ins. Perhaps you could even place the punch-in points so that they would be located just after the last "phrase", that is at the beginning of "silence". Then the reader could make sure that the pause before the rerecorded section would be exactly as long as he/she wanted. Detecting pauses (by level) is no problem. You could even set how long the pause would have to be to be considered a true pause. But inevitably sometime or another you would pause more/less than the set amount (if you speak like a normal person, that is) and with your system your system you wouldn't even be aware you'd missed it until it's too late and you'd overwritten what you wanted to keep (or be stuck with still having to edit). People have already given you the best answer for how to solve your problem. You're only hurting yourself by not at least trying what they say. |
#16
|
|||
|
|||
Ricky Hunt wrote:
Detecting pauses (by level) is no problem. You could even set how long the pause would have to be to be considered a true pause. But inevitably sometime or another you would pause more/less than the set amount (if you speak like a normal person, that is) and with your system your system you wouldn't even be aware you'd missed it until it's too late and you'd overwritten what you wanted to keep (or be stuck with still having to edit). People have already given you the best answer for how to solve your problem. You're only hurting yourself by not at least trying what they say. In fact I'm already doing what I've been suggested to do. It's not a bad way to work, but it is time consuming and I'm not convinced there isn't an easier way. I don't think pause detection would cause the problem you described. I'd be correcting over mistakes that would always be in the end of the recording, not in the middle, so overwriting would never occur. Regards, Ville Koskinen |
#17
|
|||
|
|||
"Ville Koskinen" wrote in
oups.com: In fact I'm already doing what I've been suggested to do. It's not a bad way to work, but it is time consuming and I'm not convinced there isn't an easier way. In my experience, removing the blanks doesn't take considerably more time per edit than stopping the recording, backing up, and restarting the phrase. About the only way you could speed it up significantly would be to make no mistakes in the first place. |
#18
|
|||
|
|||
|
#19
|
|||
|
|||
"Mike Rivers" wrote in message
news:znr1119008516k@trad... In article Lduse.72664$xm3.64513@attbi_s21 writes: Detecting pauses (by level) is no problem. You could even set how long the pause would have to be to be considered a true pause. And just how would you do that? Simple. Just set a threshold and how long it has to stay below that threshold to be considered a true pause. Soundforge does this already (Auto Region is what they call it). It just doesn't do it exactly like this guy wants (it won't let you "search back" a single phrase for example). Do you know of a DAW program that can detect and mark periods of silence? I know that it can be done dynamically (like the auto ID markers on a DAT or real time CD recorder) but I don't recall running into that function in a DAW program - but then I'm no walking compendium of DAW features. I don't know of any programs that do exactly what he wants either. But it's such a specific thing he wants that nobody else really seems to want so it doesn't surprise me that the feature doesn't exist. It seems that I've heard about some CD construction programs that can do that, but I can't think of which. Generally the pauses have to be longer than the pause between sentences though. Otherwise you'd get hundreds of markers where you don't want them. When I record a live show to DAT or CD, I always turn off the auto-index function and insert them manually when I can remember to do it. Right. As I said Soundforge already does this. Basically it's just like a compressor's detection circuit. It even includes attack and release. But the point you bring up about pause length is why I told him unless he speaks like a robot with some kind of precise meter and pauses of equal length (who would want to listen to that???) it's invariably going to miss a pause he wanted or jump to the wrong one. Either way he's going to have to stop and make sure he's at the right place before he continues so this way would actually take longer than the manual marker idea. |
#22
|
|||
|
|||
"Mike Rivers" wrote in message
news:znr1119026693k@trad... Simple. Just set a threshold and how long it has to stay below that threshold to be considered a true pause. Soundforge does this already (Auto Region is what they call it). It just doesn't do it exactly like this guy wants (it won't let you "search back" a single phrase for example). OK, that sounds reasonable. I didn't know that Sound Forge had that feature. Does it split the file at the point where it detects silence? You can adjust the parameters to determine where it drops a marker. Actually the "auto region" command itself just drops markers according to the parameters: attack sensitivity, release sensitivity, minimum level, minimum beat duration (how long it has to stay below treshold to be considered a pause). You can choose whether to have that marker fall on the attack of the next "phrase" or at the point when the level (release) falls on the previous phrase. Also, it either processes the whole file or you have to select what parts you want processed. If you want to actually break it up and save all the regions (phrases) as individual files you use the "Extract Region" command (all Auto Region does is drop markers). As I said before a program could be written to do what he has asked. I don't know of any that currently do it though. The programming itself wouldn't be that hard (I programmed for a living before my disability). I just don't think it would work as smoothly as he thinks it would unless he speaks with the precision of a robot (though it probably could be tweaked to work specifically for him). You would still need to manually listen to make sure you were at the right place. Personally I'd rather drop the markers manually (heck, it's just a single keypress) and go on. If so, then why couldn't he "search back" by jumping to the previous region start point? Surely it has that feature. He could but the process every time he made a mistake would be. * Stop recording. * Either process the whole file looking for regions or select just the section you want to look for pauses * Run the Auto Region command * Press the "previous region" key and then listen to make sure you are at right place. * Punch in (overwrite) and continue To me simply dropping a marker manually and continuing, then coming back and doing the clean up would be much simpler and take less time. But if he's determined to go this route (and finds a programmer to write the program for him), more power to him. I think he should edit like everybody else. There are some "time saving" features that take so much time to use that they end up being time wasters. Amen. |
#23
|
|||
|
|||
Ricky Hunt wrote:
I don't know of any programs that do exactly what he wants either. But it's such a specific thing he wants that nobody else really seems to want so it doesn't surprise me that the feature doesn't exist. I researched this a little, and some audio book authoring programs have the feature. See eg. www.daisy.org for links. I guess some organisations pay good money for the feature, which is why I'm still not convinced that the two pass recording and editing method is optimal. Unfortunately, there is nothing like this for the consumer market. Surprisingly, I haven't even found handheld voice recorders with the feature, which supposedly would be helpful for physicians and others who need to record notes. Regards, Ville Koskinen |
#24
|
|||
|
|||
|
#26
|
|||
|
|||
"Ville Koskinen" wrote in message
oups.com... Ricky Hunt wrote: I don't know of any programs that do exactly what he wants either. But it's such a specific thing he wants that nobody else really seems to want so it doesn't surprise me that the feature doesn't exist. I researched this a little, and some audio book authoring programs have the feature. See eg. www.daisy.org for links. I guess some organisations pay good money for the feature, which is why I'm still not convinced that the two pass recording and editing method is optimal. Unfortunately, there is nothing like this for the consumer market. Surprisingly, I haven't even found handheld voice recorders with the feature, which supposedly would be helpful for physicians and others who need to record notes. Regards, Ville Koskinen Ville, it wouldn't be that hard for a programmer to write you a program to do this if all you wanted was this feature. Meaning just a program that strictly records input as a WAV and does the "jump back" thing. At the end you would just save the file and then open it in another app for any other processing you wanted to do. If you ask on some of the DSP boards you might find someone who has the code laying around and might do it for a little of nothing. |
#27
|
|||
|
|||
"Mike Rivers" wrote in message
news:znr1119095241k@trad... That could be useful, but it might be pretty tedious to set up and it's bound to have some falsies, both missing things you'd like marked and marking things that don't need to be marked. I could see that in the course of a half-hour narration, in order to be useful for editing, there might be 500 or so markers - quite a forest to skip through when deciding what needs editing. Much better to rehearse so you can read it pretty well and just re-do a phrase as soon as you realize that you flubbed it. Mark that somehow, either with a marker in the program, a notation on the script, or have the producer look at his stopwatch and jot down the time. My thoughts exactly. It's one of those things that sounds simple in theory and it is easy to program. It's the voice-over artist that's the unquantifiable variable. I doubt many would want to talk in the way necessary to make it work flawlessly. And if they did I certainly wouldn't want to listen to it. |
#28
|
|||
|
|||
Ricky Hunt wrote:
Ville, it wouldn't be that hard for a programmer to write you a program to do this if all you wanted was this feature. Meaning just a program that strictly records input as a WAV and does the "jump back" thing. At the end you would just save the file and then open it in another app for any other processing you wanted to do. If you ask on some of the DSP boards you might find someone who has the code laying around and might do it for a little of nothing. I'll do that. In fact, at one point I tried to find a scriptable audio editor (like Excel VBA for audio recording) to be able to test the idea. Thanks for the responses. I've at least learned that getting the performance to a good level requires some work even if you are a professional. As a hobbyist, I guess my interest to this specific topic is something similar to a runner getting interested in GPS-navigation to measure distances and speeds (I've heard this happens). Regards, Ville Koskinen |
#29
|
|||
|
|||
"Mike Rivers" wrote:
[...] most of us here (who produce music almost exclusively) Um, *ahem*... cough cough... There happen to be more than a handful of us here for whom music is not the primary focus of our work. Part of it, sure, but not the most significant part. Some of us record people reading things, note where they've made mistakes, then go back and cut them out. Then we add music and sound effects and clips of stuff recorded in the field and mix it all up until it washes over you like a sonic tsunami! g -- "It CAN'T be too loud... some of the red lights aren't even on yet!" - Lorin David Schultz in the control room making even bad news sound good (Remove spamblock to reply) |
#30
|
|||
|
|||
"Lorin David Schultz" wrote in message
news:7uFte.71509$on1.20797@clgrps13... "Mike Rivers" wrote: [...] most of us here (who produce music almost exclusively) Um, *ahem*... cough cough... There happen to be more than a handful of us here for whom music is not the primary focus of our work. Part of it, sure, but not the most significant part. Some of us record people reading things, note where they've made mistakes, then go back and cut them out. Then we add music and sound effects and clips of stuff recorded in the field and mix it all up until it washes over you like a sonic tsunami! g Ditto here (nowadays), though I do like to read posts from you music guys. Reminds me of my early days, when the security cameras were about the most important equipment, when a well known band flew down three flights of fire-excape, when they saw a buddy of mine in his Chicago Police Seargent's uniform on the front door video monitor. Steve King |
#31
|
|||
|
|||
On or about Fri, 17 Jun 2005 13:20:33 GMT, Carey Carlan allegedly wrote:
In my experience, removing the blanks doesn't take considerably more time per edit than stopping the recording, backing up, and restarting the phrase. And then there is an advantage in having all takes there for the edit. Sometimes the reader may punch the re-start a bit too much, or rushes through the first bit of it (which they got right the first time) before settling back to the more usual pace. In these cases, the best fix may not be to just delete the first take and keep the second. It can often be better to cut from the middle of the first take into the same point of the second take. That may produce a better result than either individual take. If you try to do it with an 'automatic jump back' feature, you'd never quite be sure if it was going to jump back to the start of that sentence (or two sentences back!), the last comma, or some other point that you took a breath, that you may not have even noticed at the time that you did. The duration of each of these points compared to your set pause length will determine where it goes back to. A millisecond or two at one point or another may determine whether you still end up with some clause repeated, or else missed out because the program went back further than you thought it might. Noel Bachelor noelbachelorAT(From:_domain) Language Recordings Inc (Darwin Australia) |
Reply |
Thread Tools | |
Display Modes | |
|
|
Similar Threads | ||||
Thread | Forum | |||
on topic: we need a rec.audio.pro.ot newsgroup! | Pro Audio | |||
DNC Schedule of Events | Pro Audio |