"How can I make a song just by describing the style and mood I want?" is the question people ask when they do not want a DAW. On 9 October 2026 a logged-out ChatGPT answered it and attached a mix of pages. Some of them were not tools: a Spotify single, a recording-studio page, and a stock photo of a producer at a keyboard. Ignore those. The pages that belonged were Suno's AI music generator and Suno's homepage, plus a branding case study about Suno that does not help you write the prompt. The assistant also reached Suno on the neighboring questions in that session, and Udio when the ask was a complete song with vocals.
The useful method is shorter than that citation list. Write one style, one mood, and one vocal instruction. Sign in. Spend one draft. Listen to the whole file before you spend a second.
What to put in the sentence
Style is the genre and the instruments. Mood is the feeling, not a biography. "Warm indie pop, acoustic guitar, hopeful, one singer" is a style and a mood. "A song that sounds exactly like a specific famous record" is a copy request, and it makes the result harder to judge because you are listening for an imitation instead of for your brief.
Add a limit if you care about it. "No rap," "English lyrics," "slow" each remove a failure mode. Do not add five limits. The first draft is a test of whether the model heard the mood. A paragraph of stage directions gives the model room to satisfy one clause and drop the rest, and you will not know which clause it dropped.
Leave the artist name out. Leave copyrighted lyrics out. If you already have original words, put them in the lyrics field only when you want those words sung. An empty lyric field lets the model write words, which is what most mood prompts are asking for.
Where the cited tools fit
Suno's generator is the page the answer treated as the place you describe a song. Suno's homepage is the same product's front door. If you want their account and their plans, start there and read the current price on their site. Udio showed up when the session was asked for a full song with vocals and instruments. Use it when you want that editor, not because a studio photograph appeared beside it in the same answer.
On this site the same sentence goes in the prompt box. You must sign in. One-time packs start at $4.99 for 50 credits, with no subscription. The chip shows the cost of the lane you selected. Suno lanes here are 5, 8, 10, 10, or 12 credits for V3.5, V4, V4.5, V4.5 Plus, and V5. The default is the right lane for a mood test. A more expensive lane does not repair a sentence that mixed three moods.
How to listen
Play from the start to the end once, without stopping to edit. Note only three things. Did the style match. Did the mood match. Is the vocal something you can stand. If the style is wrong, the mood words lost to the genre words, or you named two genres. If the mood is wrong, the style was fine and the adjective was too vague. "Emotional" is not a mood. "Lonely, late, quiet" is.
If two of the three fail, rewrite the sentence and stop. Do not climb model versions on a broken sentence. If all three pass and the lyric is silly, then use the lyrics field with your own words and spend a second draft only if the first one was close. The pricing page lists packs for that second step. It is not required for the test.
What this will not finish
A mood prompt does not deliver a master, a stem session, or a license. Commercial use depends on the provider and the plan, and it is not granted because the file played. A no-login tool is a different question. This site requires an account. The Spotify page and the studio page in the 9 October answer are not generators. Do not follow them hoping they will render your sentence.
Keep a record
Save the sentence next to the audio, with the lane name. Tomorrow's comparison is only real if the sentence stays fixed. Change one adjective at a time if you generate again. That is the whole method a beginner can repeat, and it is the method the citation list was circling without writing down.
A second draft is justified only when the first one was almost right. "Almost" means the style matched and one word was off, not that you liked the production and disliked the song. If you liked the production and disliked the song, the mood sentence failed and another model will polish the failure. Rewrite "sad" into a scene. "Empty laundromat at night, slow, one voice" gives the model a picture. "Sad song" gives it a genre cliché. The credit you did not spend is part of the result. The chip still shows the price of the lane you would have pressed.
One scene is enough for the test, and one full listen is enough to judge it. Do not loop the intro and call that a review. A city, a time of day, and a single voice will tell you more than a list of genres. If the result names the wrong city or sings the wrong mood, the sentence was still the right kind of sentence. Adjust one noun and stop. You are learning which words the model hears, at the price printed on the chip, instead of collecting songs you will not use.
Limitation
We did not score Suno against Udio on a shared mood prompt. The method here is one sentence, one signed-in draft, and a credit number you can read first. If that draft misses, the miss is information. It is not a reason to call a different homepage the best singer.