But first… a quick recap and summary of the project plan.

Poems for a Rainy Day, my audio project, is a cozy poetry reading of three motivational poems: Desiderata by Max Ehrmann, Wild Geese by Mary Oliver, and Everything is Waiting for You by David Whyte. They are prose and free-verse poems that have themes on the beauty of life and the universe, with inspirational messages that everyone will probably need at some point in their life. For reference, here are the texts of the poems once again:

In conjunction with the vocal narration are ambience tracks to curate a cozy atmosphere, as well as sound effects of self-care rituals. I aim to make the output to be an encouraging pick-me-up for those who are feeling down or lost, on a rainy day. Initially inspired by my appreciation for Desiderata, it has since expanded to be a collective reading of multiple poems to fulfil the project requirements. I planned to produce four tracks in total, which are:

1. Vocal – a narration of the poetry recorded by myself

2. Ambient sounds – a recording of (1) rainfall sounds, and (2) fireplace crackling sounds

3. Foley – sound effect recordings of self-care rituals like making tea and lighting candles

4. Background music – synthesized music that I will produce using virtual instruments.

Planned to be accomplished throughout the month of August, the plan was to record the vocals, foley effects, and ambient tracks on my Maono AU-PM461TR USB microphone stationed on a DIY recording booth with a storage box lined with a blanket. The background music was to be played on the crystal synth virtual instrument from midi.city. The DAW to be used was the Fairlight module on DaVinci Resolve.  

So, what changed from the plan?

In terms of the project concept, the plan remained. The changes were more on the logistic end of things.

1. Making a more sophisticated recording booth. I decided to elevate the DIY booth with a blanket that I was going to use. I’ll detail my process in the later sections; it’s still DIY, but definitely more involved. The TL;DR of my reasoning is that since I plan to commit to doing more voiceover work in the future for my creative projects, I might as well make something that’s dedicated to this purpose, rather than assembling and disassembling the prior setup again every time I have to record something. I figured that now is the perfect time to make this. I just had to push back some of my recording dates a bit, but it didn’t come at the cost of finishing the project.

2. The process of producing my background music. As I’ve mentioned in my project plan before, I opted for a more involved way of having background music rather than simply finding some royalty-free options online. Having no musical instruments to my disposal, I arrived at virtual instruments, specifically midi.city, to record music. I planned to record covers, where I was picking between a couple of songs by Björk or anything from the Stardew Valley soundtrack. Well… as it turns out, the process wasn’t as simple as I was expecting. Once again, I’ll detail the entire situation in the production part, but the TL;DR was that I changed plans – I did cover those songs I was planning on, but I instead produced the background music through a combination of Online Sequencer to make a MIDI file, and LMMS to play around with virtual instrument plugins.

Other than these two changes, the whole project went according to plan.

The following sections detail all the steps I went through from scratch to final output. As a recap, here is a list of my gear:

• Microphone Maono AU-PM461TR USB Condenser Microphone (cardioid pattern)

• Pop filter ONAIR Mic Screen Pop Filter

• Speakers Edifier X100B PC Speaker System

• PC Intel Core i7-12700 PC build (with Realtek ALC897 as the integrated audio)

with the signal flows:

• signal inflow being the inputs (voice, foley, and ambience) going through the USB mic, then to the DAW

• signal outflow being the sound going from the DAW, to the PC's integrated audio, then to the speakers.

At the time I finished writing and turned in the pre-production project plan, I was still sick with the flu, my voice sounding very congested - and there was no way I can record a decent vocal track with it. So, I planned to tackle the elephant in the room first, which was producing my background music. I first decided which exact songs I will be covering first, in which I figured that the best option was to do one song per poem, the end and start of one indicating the beginning of a new poem. I was just playing through my choices, seeing which song fit the mood of each poem, until I have arrived at three pairings:

1. Frosti by Bjork for Wild Geese

2. Ancient by ConcernedApe (from Stardew Valley) for Desiderata

3. Ghost Synth by ConcernedApe (from Stardew Valley) for Everything is Waiting for You

I then looked up piano sheets for these songs online, which I was able to find on the MuseScore website. Here are the specific ones that I used:

1. Frosti by al.ex_p

2. Ancient  by Aubrianna Shields

3. Ghost Synth by Man With Hat

I can’t read music at all. I decided to go with Ghost Synth first since it was on the simpler side. Initially following my written plan, I went to midi.city, and using a cheat sheet for piano notes, I practiced the first few measures. When I tried to record the first few sections, I realized that... midi.city doesn’t have a recording function, it’s simply a playable virtual instrument. My workaround was to do loopback recording (recording the sounds coming out of the computer itself) on Audacity. I definitely could’ve gotten away with it by playing a little at a time, but I also struggled a lot with getting the timing right. And then attempting to sequence these short recordings on software like Audacity or Resolve – the process just became too unwieldy. I probably could have made it work with a lot of extra effort, but this was also when I started to explore other options of producing my music.

That’s when I remembered one of the discussion posts on Unit 3 by Prof. on MIDI files (which was funny as the instrument I was going to use is literally called midi.city…). This was when I started to explore the world of Online Sequencer, a website where you can make these sequences then export them either as audio or as a MIDI file. It has a very intuitive, beginner-friendly interface that I was able to figure out quite quickly. So, I settled with this method for now. I suppose maybe it’s a little easier than what I had intended to do, but definitely still involved as it still took me quite some time. It took me a couple of days to finish the sequences.

I could’ve just stopped there - assigned the instruments then exported the audio file. But I did want more control over the instruments as I saw on Prof’s post about MIDI, and potentially explore finalizing the sequence on a DAW. I ended up exporting the MIDI and continued working on the music on LMMS, a DAW. It has the VeSTige plugin that makes playing around with virtual instruments easy. I can just download them and open their interface on the plugin. It was a simple import to get the MIDI file onto the software, then I had to assign instruments to actually hear something. I explored the BBC Symphony Orchestra and the soft piano LABS plugins that Prof mentioned in the MIDI post. Both of them had intuitive interfaces, and LABS had a lot of different instrument plugins to choose from easily, so the instruments I had used just came from the both of them.

Process in LMMS

1. Frosti. While the sheet I followed was for a celesta, I actually think the original Bjork song uses a music box (from her live performances). I used the music box plugin from LABS – particularly Hannah Peel’s Music Box, then doubled the track, where I used marimba from the BBC Symphony to make it sound fuller. The third audio track was extended versions of the first notes on the measures to complement the whole track and to also give it a fuller feel. For the panning, I simply followed resources online that I had looked up - the music box was at the center, marimba slightly to the right and piano slightly to the left.

2. Ancient. Ancient’s sheet was designed for four instruments. I sequenced them separately, only selecting a part of the song to be used since the original one was over 3 minutes. Although the sheet did specify which instruments were used (glockenspiel, vibraphone, xylophone, and marimba), I did play around on LMMS with different combinations (how could I not, given I have so many at my disposal…). I ended up using the soft piano in place of the marimba, “Anthem” from Glass Piano in place of the vibraphone, music box in place of the xylophone, and harp plucks in place of the glockenspiel. For the panning, the soft piano and harp were to the left, anthem piano to the right, and music box at the center. I also used the piano roll when configuring the note velocities to match the dynamics on the sheet – some of them had crescendos, decrescendos, mezzo pianos and pianissimos that I couldn’t do with Online Sequencer.

3. Ghost Synth. For this track, I also used the same instruments with Frosti for the treble staff – doubled, using marimba on one, and music box on the other. For the bass staff, I also used the soft piano for a cohesive sound. The only difference this time around was that I also doubled the marimba track, then delayed it by half a beat for an echo-delay effect, since the original song has it too.

For all three tracks, I also took creative liberties in rearranging the songs (taking out, repeating, different parts) to match the length of the poem’s narration. And after that… I finally have three usable songs for my background music. I’ve also embedded these tracks to hear fully, since they play very minimally in the whole project:

Some thoughts before I move forward... LMMS did have something of a sequencer called piano roll where I could’ve made my sequence, but I had already sequenced everything beforehand, so I only used it in configuring the piano dynamics. But I figured it would’ve made the process easier by just accomplishing everything on one software. There also was the question of why didn’t I just do the whole thing on LMMS and potentially have gotten the +10 bonus, but at that point I had already made some headway on recording and some editing on Fairlight. The timeline was that I made the sequences on Online Sequencer, then recorded everything else, then finalized the music tracks on LMMS. It might have just complicated things if I jumped ship halfway through.

This was about the time I decided I will be making a more sophisticated version of my DIY booth. I bought a bigger storage box that was dedicated for this, then repurposed an old sleeping cushion mat to line its inside walls. Simply put, I cut the cushion to size – a rectangular one to cover the sides, then some squares for the top and bottom. After that, I did some very amateur stitching on the sides to close them back up. I stuck the newly formed cushioning inside the storage box and that was it. I did the work on the box on August 19, taking around 3-4 hours, then started recording the next night.

I was already a sweaty mess after finishing up on the vocals, but I still had the foley recordings to tackle. I took my mic out of the booth and placed it atop my desk, then prepared all the materials I’ll need: a tea packet, a mug and spoon, some water to pour, a matchbox, a candle, a plastic bag of Lays chips, and a book.

I also prepared a list beforehand of everything I planned to record to keep track of the session:

1. Window closing

2. Mug on table, mug moving

3. Pouring hot water

4. Stirring

5. Opening tea box, taking out a packet, ripping tea packet

6. Opening a matchbox, striking a match

7. Lighting candle

8. Blowing

9. Flipping pages (3 times)

10. Closing a book

I just went through this list sequentially; the action being done around 4 inches to a foot away from the mic. Everything went smoothly aside from the sound effect of me setting down the mug on the table. For some reason, since the mic was also on top of the same table, the sound of setting down the mug sounded really sharp and boom-y (I suppose it had to do with the vibrations). So, my workaround was to just hold the mic myself when recording that sound. I also moved my mic closer to the window to record a sound of it closing – my idea was in the beginning, the rain starts off very loud, then becomes quieter as a window gets closed.

Once the sound effects were done, I had one last thing to do, which was the fireplace crackling ambience using the potato chip bag. I have to be honest – I had little faith in this working, but I was actually pleasantly surprised with how it initially sounded. I did find that there was a specific technique for it to be more realistic, which was to go slower when playing with the plastic. Once I figured it out, I did a couple of takes, before finally concluding the whole recording session. I finished up past midnight.

I recorded everything in one continuous track, so perhaps one of the longest tasks was to sit through the recording output of around 30-40 minutes, cutting and organizing the tracks into the different parts, so it would be easier when positioning each of them together. Other than it getting quite boring, it was quite simple, just listening to the recordings by ear and feeling for which is the best one. At this point, I don’t even cringe hearing my own voice anymore since I’ve done this so frequently now (exposure therapy?) … Cutting out the unwanted noises on the rainfall was easy – it turns out the cuts wouldn’t be obvious, so I just cut and spliced parts freely. The same thing can be said about the fireplace foley – I put together the most realistic parts. Once done, I had organized folders of the audio, sound effects, ambience, and music tracks, which made the next part easier – laying the tracks on top of each other. Here, I don’t worry about the volumes of each track just yet – the main goal was to get the timing right, the placement of each track on the timeline: pauses between each poem, length of the intro and outro, making sure the sound effects had realistic gaps in between them, when the music starts playing, those things. Once the timeline was configured, I color-coded each track for easy identification. There were five tracks here in total as I decided to treat the fireplace and rainfall ambience as two separate tracks for more control.

Then comes the editing. My process was to individually edit each track, then adjust their volumes accordingly. I started with editing the vocal, once again, since it was the focal part of the project.

The first order of business was dynamics – compression. Listening through my recording, there are definitely some parts that are louder than others – which were visible on the waveform as well. I definitely needed compression on the track to even it out, but not too much to the point where it sounded flat, since those tonal changes also help create emphasis on the words. Resolve did have presets for dialog compression which I used as a jump-off point, then altered some sliders ever so slightly. These were the settings I settled on:

• Threshold: -33.3 dB

• Ratio – 2.5:1

• Knee – 20

• Attack – 6.9ms

• Hold – 0ms

• Release – 84ms

Though the compressor made the audio a lot more even sounding, there is still room for improvement. The next process was using the equalizer. Resolve has a parametric EQ with different presets. Going through them however, I felt that none really fit my voice or made it sound the way I wanted, so I decided to take matters into my own hands and start an EQ curve from scratch. There were a lot of good resources online for EQ cheat sheets, telling which adjustments to each frequency part does to the audio, but what worked for me was simply playing around with the bands, pulling them to extremes to really hear what they did.

But first off, what exactly did my recording need? For starters, I tend to have a very nasally sounding voice, so I was hoping to lessen that. Another issue I had was that, since I recorded my track inside a box, it did sound like it was recorded in a box, so I was also thinking that EQ can help open up my voice a lot more. Those were some of my main goals, the rest of the edits to the curve was just a matter of playing around and seeing which sounded the best. This was the EQ curve that I had settled with:

• cut on the low-end for the unwanted noise

• cut between 250-500 Hz – this was the most crucial part of the curve: the raw recording sounded too “boxy” and the cut here helped open up the sound, helping the clarity of the recording.

• boost around 125-250 – the previous cut made the voice sound a little too thin and hollow, so this slight boost helped restore some of that fullness

• boost on the high end – added some minimal brightness

For the rest of the tracks, minimal to no editing was done. The only adjustments I made to the foley SFX were noise reduction, and some individual volume adjustments on the ones that sounded too loud or got lost in the final mix. For the rainfall, I applied compression to make it more even since the rain’s sound got louder and quieter throughout the recording. Then on the equalizer, I applied a cut on the 500 Hz to open up the sound a little more. For the fireplace, I just played around with different effects to make it even more realistic. I first applied some noise reduction with all sliders cranked up to the extreme. I then raised the pitch by 7 semitones to make the crackles sound even thinner. Panning was also applied to the two ambient tracks since those two are the ones that should provide the sense of space to the whole audio. The rainfall was panned slightly to the right, and the fireplace slightly to the left, to give the idea that they are coming from those respective sides of the room. The background music was kept as it is.

Listening to the final output, I was definitely happy with it. I’ve always had voiceover issues in the past – between my voice, delivery, the mic, my acoustic setup, or poor post-processing, something was definitely amiss. I think that looking back, that voice booth was a key upgrade, since I never considered acoustics and reflections before. I also remember being too close to the mic when recording. Taking the time with the EQ made a night and day difference from the raw audio. Seeing what the adjustments in the frequencies did to my voice was so helpful given the fact that the presets never worked too well on mine. I wouldn’t have known how to improve these things had I not learned these in class.

Were there things I wish I had done differently? For sure, a lot of them. As I talked about earlier, the software workflow became too unwieldy. Had I gotten more time, or put in more effort, it definitely would’ve been nice to just accomplish everything on one DAW, LMMS, since I was also doing some light music work. I think it was just a lack of planning on my part that led to that. Also… I don’t know how to feel about my vocal delivery. I regularly do voiceovers for explainers and video essay-adjacent projects, and I think it works for that, but for spoken poetry, I’m not so sure. I think there’s a weird tone of sincerity that comes with it that I just did not have in my takes, and forcing that out of me just felt unnatural since I did note that I will be speaking in a more casual, conversational tone. There also was the perfectionism aspect. It’s so hard to declare something as done when you can keep fine-tuning and fine-tuning the tracks.

Onto the good things, I do love that I explored some light music production with MIDI and sequencing. Granted, I’m only scratching the tip of the iceberg, but it was fun playing around with virtual instruments, seeing which sounded good together, etc. I also loved the cover tracks that I made for my background music. I also was decently satisfied with the vocal processing in Resolve that I just turned it into a preset for my concurrent requirements that require voiceovers as well. Two birds with one stone, I guess. 😊

That concludes my walkthrough! Reading through my self-reflection on the plan, I think I was able to accomplish my goals – to make something decent, to be more proficient in using a DAW, and of course, to have fun with it. I learned a lot about audio production in this very hands-on process, with values like resourcefulness, adaptability, and of course, tons of trust in the process. It was nice to have this full-circle moment, from being this naïve boy in high school discovering Desiderata, to years later, creating something inspired by that memory.