It started with a short voice message from Tember. Just one minute of raw lyrics sung quietly over Instagram. I listened to it once and felt that familiar pull in the chest. The kind that says this one is going to stick with me for a while.
ORIGINAL LYRICS
I don’t have a lot of friends,
not a lot of options,
and it scares me when I do,
because then there’s more to lose,
more time to bruise,
and I’m terrified of being alone.
SimpleScreenRecorder on Linux captured the playback straight from my screen. Then I dropped the recording into turboscribe.ai so the AI could pull the words out cleanly. What came back was sparse and honest. I fed those lines into Grok and asked it to expand them while protecting that fragile feeling. Five careful iterations later we landed on the final lyrics.
From there I had Grok write a detailed music prompt and sent everything over to flowmusic.app. The result was a minimal dark electronic track with restrained trip-hop rhythm. Slow tempo. Psychologically tense. No big swells. Just looping thought patterns and a voice that sounds like someone trying hard not to feel too much.
Lab note: Sometimes the best songs begin as tiny voice notes between friends and quietly demand an entire visual world in return.
Once the track existed, I knew exactly what I wanted to do with it.
Turn it into a noir murder mystery where Detective Nora Vale is investigating a death that hits too close. The victim was someone who almost made her open one of those doors she usually leaves closed.
I started by asking Grok for 18 image prompts built around the lyrics.
I requested a strong character reference for the female detective first. Then we built the full scene list. At the last minute I switched the prompts over to Midjourney format because I wanted that crisp high-contrast noir look. A reference image of the detective was used with the –cref command in Midjourney for consistency across the set.
Step 1: Capturing and expanding the seed
SimpleScreenRecorder on Linux to grab the original voice note. Turboscribe.ai to extract clean lyrics. Grok to expand and refine them across five passes until the language felt exact.
Step 2: Turning words into music
Writing the music prompt with Grok. Feeding the lyrics and prompt into flowmusic.app. Iterating until the track felt like internal pressure made audible. 65 to 80 bpm. Muted kick. Soft distorted snare. Close-mic spoken-sung delivery with delayed echoes.
Step 3: Building the visual language
Creating the Detective Nora Vale reference image. Writing 18 detailed scene prompts that follow the emotional arc of the lyrics. Converting those prompts to strict Midjourney format with consistent chiaroscuro lighting and muted teal-black grading.
Step 4: Generation and selection
Running the Midjourney prompts to produce 72 still images total. Carefully selecting the strongest 18. Then using Grok Imagine to turn each chosen still into a smooth 10-second video clip.
Step 5: Final assembly
Bringing everything into Kdenlive on Linux. Syncing the video clips to the music. Adding subtle transitions. Letting the images breathe in time with the repeating motifs and degrading loops of the track.
Tools & Creative Stack
- SimpleScreenRecorder (Linux) for capturing the voice note
- Turboscribe.ai for lyric transcription
- Grok for lyric expansion and prompt engineering
- flowmusic.app for generating the final dark electronic trip-hop track
- Midjourney for generating 72 still images
- Grok Imagine for converting the selected stills into 10-second video clips
- Kdenlive for final video editing and assembly
Discovery / Takeaway
The real lesson was how many small careful steps it takes to move from a one-minute voice note to a finished visual story. Each tool added its own texture. Turboscribe gave me clean words. Grok helped me deepen them without losing the quiet ache. Flowmusic.app turned them into sound that feels like quiet dread. Midjourney and Grok Imagine gave me the rain-soaked shadows and hesitant expressions. Kdenlive let me stitch it all together so the images feel like they are breathing with the music.
I also learned that switching from Grok Imagine prompts to Midjourney format halfway through forced me to sharpen every description. The friction made the final noir atmosphere tighter and more consistent.
Most importantly the project reminded me why I do this. One quiet voice note from a friend became a detective walking through shadows that mirror her fear of almost staying. The pixels ended up carrying the same internal pressure the song does.
TL;DR: A one-minute voice note from Tember became lyrics became dark trip-hop track became 18-scene noir murder mystery visual album. Five lyric iterations. 72 Midjourney images. 18 video clips. All assembled in Kdenlive. The friction was worth it.
In the end Detective Nora Vale walks away down that wet street carrying the same weight we all carry when something starts to feel real. I watched the finished video and smiled because the stillness was alive.
Steve Teare
video alchemist
TerminallyBored.Monster
Palouse, Washington USA