UNITED REFORMED CHURCH
Creating a recorded service
The following is a description of what I do. It isn't neccesarily the only way to do it.
- Receive order of service from Craig (or whoever)
- Locate and download recordings of hymns and songs
- Receive voice recordings from Craig, readers, family talker and pray-er etc
- Edit and adjust recordings
- Prepare order of service
- Merge recordings into files for website, phone and live services
- Upload files to server
Receive order of service from Craig
Craig normally provides the order of service around Wednesday. The first thing I do is to number the recordings that will be needed, typically like this:
- 01 Intro
- 02 Call to worship and intro to 1st hymn
- 03 First hymn
- 04 Prayer and Lord's prayer
- 05 For all the family
- 06 Intro to 2nd hymn
- 07 Second hymn
- 08 First reading
- 09 Second reading
- 10 Sermon and intro to 3rd hymn
- 11 Third hymn
- 12 Prayers of intercession
- 13 Intro to 4th hymn
- 14 Fourth hymn
- 15 Blessing
- 16 Closing music
Sound file formats
You don't need to read or understand this section if you follow my instructions but it's here so that you can understand the reasons for the way I recommend processing the files.
Digital sound recordings come in a wide variety of formats, but they can mostly be placed in one of three categories:
- Lossless compressed
- Lossy compressed
Uncompressed files render sounds by measuring air pressure typically 44100 times per second on a scale ranging from -32767 to +32768 (referred to as '16 bit'). This is the representation used for CDs, and it is regarded by most experts as capable of faithfully capturing any sound audible to the human ear. However, it generates 176kbyte of data per second for stereo sound, so files in this format are very large. Usually uncompressed files have a .wav file name.
Lossless compressed files contain the same information as uncompressed files, but are compressed into a smaller size. There are a variety of ways of doing this, but broadly they take advantage of the fact that successive samples are normally quite similar, so storing the first and then just the difference between that and successive samples takes less space. So for example, if a series of samples is 30000, 30001, 29999 it might be stored as 30000, +1, -2. 'Lossless' means that this is done in a way that the original uncompressed version can be precisely regenerated. The commonest lossless compressed file type is FLAC, with .flac file names.
Lossy compressed files are (usually) the smallest of the three types. These files drop some of the nuances of the sound that are judged to be less audible to a human ear. There are many lossy compression algorithms, and more are continually appearing. It takes a significant amount of computing power to compress and decompress the sound, and if they are used in consumer devices the device needs to be able to decompress in real time - ie as fast as the sound is played - and that can limit the efficiency of the compression. The commonest lossy format is probably MP3 (.mp3 file names) but there are many others.
If a recording is made in a lossless format and then finally compressed into MP3 at 44100Hz then few people can hear any degradation in sound quality. However if the track has been compressed into a lossy format, then decompressed for editing, and then compressed again, sound artefacts become increasingly obvious. Also, if the sampling rate is changed during editing, fidelity will always be lost, and typically the elements of our services come in at anything from 16000Hz to 48000Hz. Anything downloaded from YouTube has probably already been compressed and decompressed at least twice.
The significance of these formats is that there is loss of sound quality in lossy compression and decompression, and in editing sound files it is therefore best to minimise the number of times that happens. The incoming files for our services are usually compressed, so the process I follow is to uncompress them once, do all the editing and joining using uncompressed files, and then I compress the result for uploading.
I use a Windows 10 computer and these instructions assume that you will too. I'm sure it can all be done on a Mac, but I don't have that knowledge. I think you would struggle to do it on a tablet or phone, but if you have a Linux computer the instructions will almost work. You will need a device with speakers - preferably external.
There are a number of software tools required:
- A sound track editor
- A CD 'ripper' to copy tracks off a CD
- A YouTube downloader
- A file transfer utility to upload recordings to the website
By far the best free sound editor is Audacity. In fact, it is so good that many professionals use nothing else. You can download it from https://www.audacityteam.org. You will also need the FFmpeg library, which is listed as an 'optional download' on the same page.
There are lots of CD rippers available, but the best one I have found is Fairstars CD Ripper, which is free. Unusually, it can rip CD tracks into a flac (lossless) format.
I mainly use the 'sound fron video' tool to extract the sound track of hymns from YouTube videos. I use a tool called Viddly. I have bought the paid version as it allows better quality downloads and doesn't have ads, but the free version is probably adequate. The paid version allows AAC format downloads, which are better because that is the format used by YouTube, so no decompression/recompression step is involved. There is also an online downloader which I have used. If you use that, select the .aac format for downloading. It's worth having both of these as neither is entirely reliable.
The file transfer utility I recommend is FileZilla. You want the client rather than the server. There are many other such tools and they all do the job.
Apart from those three you will need a document editor to prepare the order of service. I use Microsoft Word, but LibreOffice will work with a few tweaks to the formatting.
Editing the sound files
I start by creating a folder for the service in my Documents area. I call it something like '20200614 Good news'. Whether from email or by download or ripping from CD, I save files in their incoming format with file names starting with the 2-digit number from the order of service - see above.
So, for example, at one point in preparing a previous service the contents of Input were:
01 Intro.wav 03 O_Breath_of_Life.aac 05 Family.m4a 07 Spirit of God, unseen as the wind.ogg 08 Corinthians.aup 09 Acts reading.m4a 11 The Spirit Lives To Set Us Free Song Lyrics Video.ogg 12 Prayers of int.m4a 14 TRENTHAM-BREATHE_ON_ME_BREATH_OF_GOD[Youtubemp3.download].aac 16 We_Are_One_in_the_Spirit[Youtubemp3.download].aac
You can see that I have a variety of different file formats, uncompressed, losslessly compressed and lossily compressed, but fortunately Audacity can read them all, decompressing where necessary. Audacity can also extract the audio from most video formats (.mp4, .mov and so on).
Item 01 is the few words that I say at the start of the service. I usually record that directly into Audacity using an external mic, but you can use other means. There is some guidance here.
All these files need to be edited, and in the next sections I've split them into the voice items and the musical ones.
Each voice file needs to be edited separately in Audacity. First, read the file into Audacity, either by right-clicking the file name in Windows Explorer, choosing 'Open with' and then 'Audacity', or you can run Audacity from the start menu and then use File → Open just as you would in an app such as Microsoft Word.
If you'd like to follow along with me as we prepare this file, right-click here and save the file somwhere convenient.
You will see a visualisation of the sound, like this:
The fuzzy blue part is a visualisation of the sound. Time runs from left to right and the scale above the trace is in seconds. The vertical scale is pressure: you can take that as meaning that the further the trace departs from the central 0.0, the louder the sound. The six round buttons towards the top left are from left to right: pause, play, stop, go to the start, go to the end and record.
This is a very short file. For longer files you will need to use ctrl-1 to expand the time axis, ctrl-2 to compress it, and ctrl-3 to return to having the whole track visible at once. If you have a mouse with a wheel, using it with shift depressed on the keyboard scrolls through the file, and with Ctrl depressed zooms in and out.
The trace shown starts and ends with around a few seconds of silence - or near silence. I edit all the voice files to have 2 seconds' silences at the start and the end (because some sound cards require it). To do that:
- Click just to the left of where the sound starts - at about 2.3 seconds
- Hold down the shift key and type 'j'. That selects everything from where you just clicked back to the start.
- Press the 'Delete' or 'Del' key on your keyboard. That will delete the selected silence.
- Go to the Generate menu, and select Silence. A box will pop up headed 'Duration'.
- Set the duration to 2.000 seconds and click on OK.
- Now click just to the right of the end of the sound, hold down shift and type 'k'. That will select from where you clicked to the end.
- Repeat the deletion and insertion of silence.
If you're following along, click the green 'play' button to hear the sound (assuming you have a computer with speakers).
There are then a few things potentially needed. The first is that some recordings are nominally in stereo. Those will appear as two traces stacked vertically. If that's what you see, type ctrl-A to select both tracks and then go to the Tracks menu, select Mix and then 'Stereo track down to mono'. That will combine the two tracks into one.
There are two other adjustments you may need to make. The first is to adjust the 'equalisation', usually called EQ, and the second is to adjust the overall volume. They should be done in that order, but let's look at the second one first.
Sound level is measured in decibels (dB). The definition is complicated, but roughly, amplifying by 5dB makes sounds 'a little louder', by 10db 'quite a bit louder' and by 20db 'a lot louder'. Our aim is to make all the speakers in a service sound about the same loudness. The only way to do that is to listen and compare, as, strangely, the traces of sounds which sound about equally loud often look quite different on the screen.
The trace you have in front of you is me speaking, and needs to be amplified by 3 or 4 dB. Craig's recordings are often rather quieter and need to be amplified a little more - usually 5dB is about right. Other speakers vary: some sound over-loud and others need amplification. I play each recording with the volume control on my speakers at the same level and judge whether they need to be amplified or reduced. It isn't an exact science.
To do it first select the entire track. The easiest way is to hold down your Ctrl key and type
'a' (for 'all'). Now go to the Effect menu and select Amplify. You will see this box:
The Amplification shown (9.14 dB) is the maximum anount of amplification you can apply without 'clipping'. Clipping means that you have tried to amplify something in the recording more than the digital representation can accommodate - which probably equates to too loud. If you do clip the sound, it will probably sound nasty, so don't normally do it. However, occasionally the clipping is just a momentary event which will be unnoticeable.
In other instances you may be able to delete the offending sound - often appropriate if it's caused for example by the speaker touching the mic with his papers. You'll see that Audacity doesn't allow clipping unless you deliberately tick the 'Allow clipping' box.
Normally, it isn't necessary to amplify more than 5-8 dB. Occasionally, speakers are too loud, and you may have to reduce them by choosing a negative amplification, say -5dB.
If you're trying different amplification levels it's best to return to the original level: so for example try 3dB and if it isn't enough, type ctrl-Z to undo that and then try 5dB.
EQ is an adjustment whereby different frequencies are amplified by different amounts, usually to compensate for low-fidelity microphones. Audacity has some very advanced capabilities, but unless you're an experienced audio tech I suggest that you only consider the basic Bass and Treble control in the Effect menu, and then only if the recording sounds unnaturally shrill (too much treble) or rumbly (too much bass). Typically I would increase or reduce by 3db (eg increase treble by 3dB and reduce bass by 3dB).
Another effect I sometimes use is Noise Reduction. If you have background noise in what should be silent parts of the track, select a sample of that, go to Effect → Noise Reduction and click on Get Noise Profile. Then select the entire track, go to Noise Reduction again and it will be filtered out. This is often needed if you have had to amplify a track by more than about 8dB.
When you've done all that, go to File → Export. If Export as FLAC appears as an option, choose that; otherwise choose Export Audio ... and then select FLAC. If this is the final version, save it in the Final subfolder.
This is the most time-consuming part of the exercise - unless we have the song on file already.
Ones we have already are documented in the Songs database. Log on to the website, go
to Worship and then Songs & services. An olive green blob against a song title means we
have a downloaded song, and a violet blob means we have it on a CD. If you click on, for example,
All my hope on God is founded and then scroll down below the lyrics you will see this:
This shows that we have one downloaded recording. You can listen to it by clicking on the arrowhead on the 'Server file' line, which will play an mp3 copy (stored on the webserver) of the better quality flac file (stored on my home server). If the track had been downloaded from the internet (eg YouTube) there would be a link in the 'Source URL' line, also with a play arrowhead.
Also, at the bottom of the page, you can see that we have it on a CD. Some of these CDs are quite good, others less so, but they are worth checking.
For each song where we don't already have a recording, go to YouTube and search for it. You are looking for a version where you can hear the words clearly, which could be a congregational version or a soloist who doesn't add too much variation in notes, words or tempo. BBC Songs of Praise recordings are often good from a musical viewpoint, although many have been uploaded by means which reduce the technical quality. Some videos have on-screen subtitles or lyrics in the comments which is a definite plus, as otherwise you have to listen carefully to what is being sung in order to produce the order of service. However, check that the lyrics are correct! Note too, that the number of verses or the order of the verses may differ from what we are used to.
Be a little sensitive to the source: I have avoided the Mormon Tabernacle Choir although I'd probably use them if theirs was the only usable recording.
If there is no suitable recording available, consult Craig. He may have an alternative choice of hymn.
Once you have chosen a recording, make a note of who is singing (sometimes you have to click on SEE MORE in YouTube). Download the sound using one of the tools mentioned above. For the free version of YouTube Downloader you are stuck with 128kb/s* MP3, but with the paid version you can download 192kb/s AAC, which is better. The website downloader also offers .aac, which is a good choice because that's what YouTube uses and although it's being decompressed and recompressed as it downloads, at least you are not introducing any new artefacts.
You then need to edit the file in the same way as voice tracks. People expect music to be louder than a single voice speaking, so you can get away with a louder overall volume. Sometimes I have had to boost a quiet verse or tone down a loud verse to get an acceptable whole. Some EQ-ing of the track can help, but other than suppressing over-enthusiastic organ bass lines I tend to be sparing. I generally trim off any initial playover of the tune if it's a well-known one.
It has become my habit to include a music track after the blessing. I try to find something relevant to the reading or sermon topic, but it isn't always possible. I try to balance modern and traditional tracks and also try to match the mood of the service - something solemn for Good Friday and triumphant for Easter Day.
Putting it all together
I have not found a easy way to concatenate a number of audio tracks. What you have to do is to open track 01 and skip to the end (ctrl-A to select all and then K to go to the end). Set the project rate (bottom left of the window) to 16000. Now open track 02, which will open in a separate window. Press ctrl-A (select all), ctrl-C (copy), switch back to the first window, type ctrl-V (paste) and then K.
Now go back to the track 02 window, type ctrl-W (close), ctrl-O (that's letter O for open) and open track 03.
Now repeat the select all, copy, switch window, paste, skip to end, switch window, close, open next track and so on.
When you've pasted all the tracks together, export the file as 2020mmdd_ws.flac in your Final folder, where mmdd aret the numeric month and day of the service (not today!). So for example 20201225 is Christmas Day. I then go through the service listening especially to the joins between the original recordings and checking against the order of service to make sure everything is there. You'll see black vertical lines in the trace where the joins are, so it's easy to find them. This may also reveal segments that are too loud or soft, and you can select portions of the recording and use the Effect → Amplify tool to fix that.
Once you are happy with the result, export the file again.
Now go to File → Export audio and choose 'save as type' to be MP3. Bit Rate Mode should be set to Preset, Quality to Medium and Variable Speed to Standard. Save the file as 2020mmdd_ws.mp3.
The telephone service provider can only handle tracks up to 40 minutes long, so a service usuallly has to be split into 2 files. The best place to split is usually just before the sermon. So find that place in the trace, click in the silence and type shift-J to select from the start to that point. Go to the File menu and this time choose Export Selected Audio, and save the file as before as 2020mmdd_phone1.mp3.
Now click again in the middle of the pre-sermon silence and this time type shift-K to select from there to the end. Export that selection as 2020mmdd_phone2.mp3.
These exact file names are critical to the automatic deployment on Sunday morning. For a service at another time of day, say 4pm, add the time like this: 20201224_1600_ws.mp3.
An added complication arises when the recorsings are being used in church with some segments are being given by live speakers. Briefly:
- When concatenating the files, after each one click back-arrow, ctrl-B to open a label track.
- Type in the name of the track you have just pasted (eg Sermon)
- Type return, up-arrow, K to place the cursor at the end of the merged tracks.
- ... and repeat for each track.
It is now relatively easy to select each set of segments to export separately as mp3 files for each church. Often you will end up with 15 or so output files and it is wise to check that each contains what you think it does.
Order of service
I've left this till now, but in reality I find it best to do in parallel with the audio files, especially the hymns. You can download a sample OOS here. As previously mentioned you will need to spend some time making sure the words match the recordings.
If you have access to the St Andrew's songs database you can find the words of most hymns there, which is a good starting point. Craig, Heather Saint, Richard Seaton, Janette Walls and Rosemairi Evison have access.
Once you have finalised the OOS, save it in both the original (eg Word) and PDF format. The PDF version should be named 2020mmdd_oos.pdf where mmdd is the date of the service.
Uploading to the server
Open FileZilla. Click on the Site manager button, which is just below the word 'File' - I've outlined it in red:
Then click on New Site and give the new site a name such as St Andrew's. On the right-hand side, select and complete the fields as shown above. The password is genesis1525 . Once you've done that, click on OK. You only have to do this once: it will remember these settings for next time.
Now click on the down-arrow symbol just to the right of the Site Manager button you used earlier. Select St Andrew's. You should now see something like this:
You only need to do the above once: it will be there for you the next time.
The panel on the left shows files on your computer. The panel on the right shows those on the webserver.
Start by finding the files on your computer. You can type the folder location in where it says 'Local site' - it might be something like C:\Users\Fred\Documents\Recordings\20200706 Good News\Final.
Then find the right folder on the server, which is public_html/audio. If you've found the folder it will be full of audio files, as shown.
Once you have the two folders aligned, double-click on each of the three mp3 files you created:
It will take a few seconds or even minutes depending on the speed of your internet connection.
Now upload the OOS. On the right-hand side you need the folder /docs/oos. Your OOS file
Once you can see all the files on the server you can close FileZilla.
The next part will need to be done by somebody who has write permission to the St Andrew's
songs database. Currently these are:
Peter Campbell Smith
Steffi Campbell Smith
Christy and Richard also have the permission to add to that list.
The person with that permission needs to go to the St Andrew's website. log on, and then go to 'Worship', 'Songs and services' and 'Services'. Normally, the displayed service will be the next one to take place, but if not the links to 'preceding service' and 'following service' will get there.
If all is well, the names of the uploaded files will appear under 'Audio files'. You can click on each of the files to hear them and check they are the right ones.
The processes to make these files live on the website and the phone service provider at the appropriate time should now happen automatically.
* There are 3 measures used to indicate the amount of data used to represent audio: bits per sample (eg 16-bit), samples per second (eg 44100Hz), and data rate (eg 128kb/s). The relationship between these in uncompressed audio is that bits per sample x samples per second = data rate, ie bits per second. Lossless compression (eg FLAC) will reduce the data rate by typically 30-50%, lossy compression (eg MP3) can do so up to 80%.