ONE · WHAT SHIPPED
Suno released Studio 2.0 on the thirteenth. MIDI is the headline: you can import it, record it, and edit it on the timeline, and there is a wavetable synth for designing your own sounds. If you have no controller, you can play notes from your typing keyboard. The part worth stopping on is smaller than the announcement and easy to miss, so here it is on its own line. A MIDI clip can be used as the prompt for a new audio generation.
The rest is a studio catching up with what a studio is. There is a chat bar in beta that will make instruments and vocals, design plugins, and tidy a session. Stem separation was upgraded, so outside audio can come in, be split, and be rearranged. There are factory effects including sidechain compression and convolution reverb, and you can have the chat build plugins that do not exist yet. Automation curves arrived, so parameters can move over time.
None of that is what changes the work. What changes the work is that the thing you hand the model no longer has to be a sentence.
GO TO THE SOURCE
Suno's own write-up is here. Read past the feature list to the section on what they learned from artists, which is the only part that argues with itself.
TWO · THE CEILING THE VENDOR ADMITTED
Underneath the features, Suno wrote something a company selling text-to-music has no reason to write. Text prompts, they said, work well for finding inspiration, and are no substitute for playing a part in with a keyboard or recording it with a microphone.
That is worth sitting with, because most of us have spent two years getting better at the thing they just described as the weaker half. We have learned to write longer, to stack adjectives, to name three artists and a decade and a room. It produces something. It rarely produces the thing, and the gap between those two has been quietly blamed on the model.
The honest reading is that words are a lossy way to carry a musical idea, and they always were. A four bar line played badly contains information that a paragraph written well does not. Pitch, interval, timing, where the accent sits, what does not happen. You can hum it in ten seconds. Describing the same ten seconds accurately takes a page and still arrives approximate.
THREE · WHAT IT COSTS, PLAINLY
Three limits, none of them hidden, all of them worth knowing before you plan around this.
Exporting multitracks and stems at 32 bit and 48 kilohertz without limits is a Premier subscription thing. If you are on a lower tier, the studio is still there, but the door you walk your finished work out through is narrower. Building your own plugins from the chat bar draws no credits at launch, and Suno says plainly that a credit structure may arrive later, so treat the current arrangement as a window rather than a settled price. And the chat bar is in beta, which is the word that means the behaviour you learn this month may not be the behaviour next month.
None of that is a reason to skip it. It is a reason to put your weight on the parts that are yours, which are the MIDI you played and the audio you recorded, and to keep those somewhere that is not one company's timeline.
PROMPT OF THE DAY
The one that makes the chat bar describe your arrangement back to you.
Use it on a session that is nearly there and will not sit right. Paste it into the chat bar with the session open.
Describe this arrangement back to me as it currently stands. Tell me, separately: what is holding the low end, what is carrying the pulse, what is carrying the melody, what is filling the middle, and which frequency range is crowded or empty. Then name the one element that is missing, and say what job it needs to do rather than which instrument should do it. Do not add anything. Do not change anything. Do not tell me it sounds good.
The last three lines are the ones doing the work. Asked without them, a model will improve your session and hand it back, which teaches you nothing and costs you the version you had. Asked with them, you get a map, and the missing element is usually something you already suspected and had been calling a mix problem.
THE NEXT WORKING NOTE
The habit this suggests costs nothing and does not depend on Suno. When an idea arrives, record it badly instead of describing it well. Four bars, typing keyboard, phone microphone, whatever is nearest. The recording is a worse artefact than your description and a far better prompt, and it stays useful whichever tool you are in next year. And when the track is finished, the harder problem starts, which is getting anyone to hear it: Suno has a short piece on making short-form content for your songs that is more practical than most.
I am putting together a motif sheet: short four bar starting points in a range of feels, written out as note names so you can type them in without a controller, each with a line on what it is good for and what it fights with. Reply with send me the motif sheet and I will get it to you when it is done.
A QUESTION FOR YOU
When something you made came out close but not right, was the idea actually in the prompt. I keep finding mine was in my head and never made it into the words, and I had been blaming the model for not reading them.
| 
