AI Music Production Workflow
From AI Song to Streaming Hit
Welcome to the world of AI music production! Here I'll show you how we at nuonu ("NEW ON YOU") take a song from idea to worldwide release.
1. The Song Idea
-
Classic or via AI: ChatGPT & Co. are your friends here.
-
Structure Lyrics: Verses, Chorus, Bridge – no AI phrases, please! Your text lives, is adapted, and trimmed to the song structure.
2. Create Song
-
Studio or AI: Tools like Suno and Udio make it possible to create your songs completely online.
-
Tip: Make sure the AI sings exactly what you've written. A little trick: with „extend“ you can continue or end songs.
-
Software Choice: Suno delivers creative results, Udio provides cleaner productions. Version 4 of Suno has been available since late 2024, with new voices and instrumentations. But be aware of potential shimmering and bloated instrumentals! I personally only work with Suno and have created almost all the titles shown here in V4 or subjected them to remastering.
- Export: Get whole Song delivers the finished song, provided it consists of several parts; afterwards, we only need the WAV.
3. Lyrics
- Check the vocals and save them in a .txt file, and save them with the song. We need these for Suno, YouTube, mp3Tag.app, and for SEO.
3. Mastering
-
Turn up the volume: Radio-ready LUFS values (12-14) are the goal. Logic or Isotope RX11 help with this.
-
Use AI plug-ins: Remove noise, shape the lows, and make the sound present. The songs are already very good since V4, and standard mastering in Logic Pro is usually sufficient.
-
New since 2024: Suno allows instrumental and vocal to be saved separately. Perfect for finetuning! 2 stems are generated, but usually not needed. It is often more sensible to create several remasters and then cut them in the DAW.
4. WAV and MP3 Export
-
Tips: Leave space at the beginning and end, avoid dithering. Raise the level to 0 and check that no track is overdriven.
5. Create Cover Images
-
AI or Classic: Tools like Midjourney, DALL-E, or Firefly deliver creative results. For videos, you can generate a series of images.
6. Upscaling
-
Sharper is Better: Upscale with Topaz Gigapixel or Lightroom to get the most out of your images.
7. Define Band Name & Title
-
Logo-Ready: Your band name as PNG saves time for further projects.
8. Design Album Cover
-
Mind the Format: Square covers (3840 px x 3840 px) are standard; remember that including images in the MP3 Exif data unnecessarily bloats the file, and Suno doesn't like files that are too large. An ideal size here is 1080 px x 1080 px. A web export often suffices at 15% quality, meaning just enough so you can barely see any artifacts. The size is usually reduced to around 100 KB. Keep in mind that while there are better formats like AVIF or WebP, not all players support them.
9. Add MP3 Tags
-
Enter Metadata: Title, artist, album artist, year, genre, album cover, etc. – indispensable for end devices. Mark the final version (e.g., „Master“) or do as I do and prefix completed projects with an underscore (_). I save finished projects directly into WAV and MP3 folders, so projects can be completed. I do all this
-
with the mp3Tag.app and also insert the lyrics or, for instrumentals, the content directly into the comments there.
10. Release as .wav and .mp3
-
All Platforms: Use Feiyr for streaming and sales. Good support and great handling! But there are many other distribution platforms as well.
- Own Website: Consider whether it's not better to present your own offering. I created this one after many tests with this platform.
11. File Control
If you have considered everything, then you have the following files:
Project Data
- nuonu - Title.wav (At least one original WAV from your music AI, e.g., Suno, Udio, Recording)
- nuonu - Title.txt (Lyrics or for instrumental: bpm, moods, instruments...
- Midjourney or other original image file, probably as .png
- Midjourney or other image file upscaled to 3840 px.
- nuonu - Title.psd (Photoshop file of your cover)
- nuonu - Title.jpg (Album Cover in full resolution)
- nuonu---Title-1080.jpg (Album Cover compressed)
Central Data (each as a collection in a folder):
- Logo as .png (Outside the audio project, as it is needed for every project)
- nuonu - Title.mp4 (if you have also created a video)
- nuonu - Title.mp3 (Final Master incl. mp3Tag metadata)
- nuonu - Title.wav (Final Master incl. mp3Tag metadata)
Since there can be different music genres depending on interest, I have again categorized the MP3 and WAV folders, as they are also listed here under Albums.
12. Release as .wav and .mp3
-
All Platforms: Use Feiyr for streaming and sales. Good support and great handling! But there are many other distribution platforms as well.
13. Create Music Video
-
Social Media Friendly: Square, fast videos in cover format often perform better than elaborate productions.
14. Social Media
-
Tips on Lyrics: YouTube automates subtitles, but manual looks simply better.
-
YouTube, Facebook, Instagram: Here you now need the previously created files .txt, .jpg, .mp4...
15. Automation
As a final step, I needed a website like nuonu.com and a player, as well as my own plugins to automate publishing, sharing, and distribution.
The experiences are now flowing into plugin development. This currently consists of 2 plugins with automation processes.
audio-folder-player-pro-json
- Creation of an album-specific .json database based on getID3.
- Creation of zip downloads of entire albums.
- Creation of -40, -512, -1080px sized cover images from the metadata.
- Player with its own gesture control, which also allows listening to tracks on different devices.
- Asynchronous filling of IDs on the page based on the .json data.
- A sharing module that allows sharing a song, but presenting it together with the album.
audio-folder-player-list
- A plugin that presents and links .json data (visible here under Albums).
All this greatly simplifies management and automatically ensures that all MP3, WAV, cover, TXT... data are backed up online.
Both plugins are tested live here and can also be purchased if interested.
Costs
-
Software (Logic, Isotope, Topaz, Photoshop, Midjourney, ChatGPT etc.): approx. 30 to 100 EUR monthly.
Conclusion
AI music production opens doors for everyone – whether musicians or beginners. Depending on the project, you will work on a song for 1 (short instrumental without cover) to 8 hours (with your own lyrics and a bit of attention to detail). You don't need to be able to sing or play an instrument and can still be creative. Ideal for background music and creative experiments. Live performances remain reserved for classical musicians for now, but who knows where the journey will go? Perhaps we will soon experience the first AI world tour.
Stay tuned, stay creative – and let AI work for you!
