What I learned from this video poem?
Everything seemed cohesively unrelated during the “build” – is that chaos? Until – at the very end – I dedicated the work to my Mom. Then it all made sense to me. She was a 1960s work-at-home Mom with two little kids. She married an unpredictable raging scientist and narcissist. She was the unwilling victim of domestic violence. She chose never speaking about the bullying or violence. Silenced by intimidation and weakness. But to her credit, she survived marriage but could never escape mortal death. My child-mind thought I could protect her from death. But I was wrong. She died at 82 from complications related to age.
What a horrible way to live. Cowering and manipulating each other. Walking on eggshells.
Don’t Call Me Dear is a poem written and self-published in print during the early 1990s. The original piece was biographical and about a guy (me). I was divorced at the time. To mask my vulnerability, I switched the gender to female – a common creative-writing trick. Often in my writings, the female character represents my creative inner self.
The goal of lo-fi video poetry is working with no costs. Severe self-imposed limitations for productivity and creativity. New ideas emerge. Serendipity helps. The goal is finishing and moving on to the subsequent project. There’s no perfectionist fine tuning after final output. Acquired learning only translates forward to new projects. Not to reworking past flaws. The flaws are part of the prolific process.
In hindsight, there are things I’d do differently. But that’s not a disciplined option to stay on course. Quality is sacrificed – but successive projects improve and gain new skills. Future achievement grows from adaptive practice. You acquire new habits and skills.
Ashamed, almost terrified she listened.
Panic swept her frozen face.
With whirling mind she thought,
Don’t say those words.
Please, please don’t make me say it to you.
In a flash, she recalled every hurt and pain received under the umbrella of love.
Searing tears still smoking in her tender heart.
“Can I believe you?”
“Can I trust you?”
It still stings so deep
You reopened all my old wounds.
Do you understand the real me?
The fire smoulders now.
Bits of ash flutter off in the breeze.
The pungent smell of burnt memories spin slowly in the billowing smoke.
Tomorrow is gone but yesterdays feelings live on.
Will they ever sleep?
They toss and turn.
Moulding dreams haunted by fears without meaning.
Don’t call me Dear
I then salvaged a piece of a 2017 letter I wrote to a “terminally-gifted,” 17-year-old girl. She’s a brilliant friend of Levi, my son. I asked her to pose any question to me. What would she choose? The question she asked, “How do boys think?” That response was a 1-page letter my son Levi later delivered to her. I excerpted a few lines for this project as a voice test sample. In the end, it fit adding the excerpt to this piece.
Testing Robot Voice Talent from Amazon
There are four contenders racing to capture market share in Artificial Intelligence (AI) text-to-voice and voice recognition. Those are:
Each has different goals and features and are incorporating these into the programming language called Speech Synthesis Markup Language (SSML). Learn more about SSML
To generate two voice-over tracks, I used the online text-to-speech synthesizer on Amazon Polly. A sign-up is required. Then you access a development tool that processes up to 3,000 characters. They charged me $1 via PayPal. You can listen to the text-to-speech results before download as an MP3 file. It feels like a live environment compared to other cumbersome techniques.
This production method was infinitely better than the clunky workaround using SpeechKit WordPress plugin. Before, I couldn’t preview nor place timed gaps between sentences. Very cumbersome. That’s how I produced my first two very-novice and mechanical robot-voiced videos, Last Rocket and Sanctum Nova Muerte. Still, they were better than even older alternatives.
Converting Talent-less Text-to-Voice
The first text-to-voice track built for this lofi project is an Amazon-Polly Kendra conversion of the letter excerpt. The second MP3 is the same text read by Matthew.
REFERENCE: Voice Samples
What’s SSML? Who cares?
The Speech Synthesis Markup Language (SSML) is used with a standard Application Programming Interface (API). SSML is used to control the volume, pitch, rate, and pronunciation of synthesized speech. Using SSML, I coded variable-length pauses without slicing or splicing the tracks into pieces to insert silent gaps. This was a big step forward for productivity. Ideally, and eventually, future voices will switch/change during output instead of editing in a production audio editor. But not yet. There’s a provision for specifying preferred voice in SSML code – but it doesn’t work yet. You choose one voice in the Amazon Polly dashboard – not in the SSML code. That’s it. One vice per track. Oh, well. Soon. Maybe?
I combined the two MP3 tracks (voices) and edited using Audacity on a Linux operating system. My computer is built from scrap, secondhand, used parts purchased on Ebay for under a few hundred dollars. That price tag includes three LED back-lit monitors. That is cheaper than hiring professional voice talent.
Polly Voice Track Output Examples
Here are the sample MP3 Kendra and Mathew voice tracks and then the composite follows:
Kendra solo track above
Mathew solo track above
Above: Composite of two tracks in Audacity.
Audacity is a free and open-source digital audio editor and recording application software available for Windows, macOS/OS X and Linux operating systems.
Another female voice track.
I then proccessed the 21-line poem Don’t Call Me Dear using speech synthesis techniques – but with Polly’s Salli voice. Here I added more effects – besides breaks or gaps. I used breath duration, whispered, and break time settings in the SSML code. That 1-minute and 17-second code example is below. It’s downloaded from Amazon’s dashboard.
Still Image Selection
Since there are 21 lines in the poem, it takes about the same number of still images to create a slideshow. Each line gives some clue how to communicate feelings and ideas. I rename and number the downloaded images for automatic sorting when dropped into OpenShot video editor.
I used Google Image Search and creative-commons to locate qualified large images. They were not resized or optimized for speed after download. They are processed in free PhotoFilmStrip software on Linux to match the duration of the final voice track.
Video Compression saves time during upload and page loads.
After the composite video is output from OpenShot video editor, the MP4 video weighs 383MB. It’s optimized using HandBrake software for translation to web optimization and conversion to M4V format.
HandBrake changes the file size to 48MB. This is important for an 8-times faster upload to my YouTube channel. It only took about 10 minutes uploading.
21 lines = 21 images.
Then off searching the Internet to harvest open-source image based on keywords. A lot like hunting for the right driftwood washed up on sandy shores.
- Ashamed, almost terrified she listened.
- Panic swept her frozen face.
- With whirling mind she thought,
- Don’t say those words.
- Please, please don’t make me say it to you.
- In a flash, she recalled every hurt and pain received under the umbrella of love.
- Searing tears still smoking in her tender heart.
- Can I believe you?
- Can I trust you?
- It still stings so deep
- You reopened all my old wounds.
- Do you understand the real me?
- The fire smoulders now.
- Bits of ash flutter off in the breeze.
- The pungent smell of burnt memories spin slowly in the billowing smoke.
- Tomorrow is gone but yesterdays feelings live on.
- Will they ever sleep?
- They toss and turn.
- Dancing insomniacs.
- Moulding dreams haunted by fears without meaning.
- Don’t call me Dear.
Several extra images were discovered and added for a total of 27 images. Some were later deleted and substituted.
What is Stock Video?
Stock Videos are videos that are about 60 sec made for commercial purposes. These videos are sold on photo stock websites for commercial purposes. Usually they are used on TV, Internet advertising, or for interface design. We hunt for the freebies or promotional giveaways.
Sources of free video clips: