Text to song prompts work best when they treat a 10-second clip as a focused test, not a miniature finished single. This set checks whether ACE-Step v1.5 can hold onto a clear genre, instrumentation, vocal instruction, and short lyric structure across six very different briefs. Every player below contains the original 10-second output already uploaded to this blog. The goal is practical: learn which prompt details survive a short generation and which should be simplified before making a longer track.
What this test set checks
All six examples use Text To Song (ACE-Step v1.5) on Wiro with a requested duration of 10 seconds. The test varies genre, energy, vocal use, and whether lyrics are supplied. It does not test reference audio, cover generation, repainting, or longer arrangements. That matters: a convincing 10-second opening does not prove that a model will sustain a full verse, chorus, or mix transition.
The Wiro model interface exposes caption, lyrics, genre, instrument, mood, vocal settings, duration, language, BPM, inference steps, guidance scale, planning, instrumental mode, and seed. In these examples, the written caption carries the main style request; the 10-second duration is the constant. Vocal examples include short structured lyrics. Instrumental examples explicitly use [Instrumental]. The original post did not record seed, BPM, steps, scale, or a reference-audio input for each clip, so those values are not retroactively claimed here.
The available Wiro documentation lists 10, 30, 60, 90, 120, 180, and 240 seconds as duration choices. It does not publish a Wiro-specific elapsed time or price for these six saved outputs, and no new paid runs were made for this update. There is therefore no honest per-output Wiro cost or runtime to report. The ACE-Step project reports hardware benchmarks of under two seconds for a full song on an A100 and under 10 seconds on an RTX 3090; those are project benchmarks, not a promise of Wiro queue or render time.
Text to song prompt results
1. Synthwave pop hook
Caption: Upbeat synthwave track with driving bass, retro drums, shimmering pads, and catchy female vocal hook
Lyrics: A two-line verse and a repeated chorus hook about neon lights and riding the wave.
This is the right kind of brief for a short vocal test: it names the rhythmic engine, the period color, the main layers, and the vocal role. The repeated chorus phrase gives the model a compact target rather than a full narrative. Use this shape when the clip needs to sell a hook or a campaign sting. For a longer song, add arrangement sections and leave room for the lyric cadence to develop.
2. Lo-fi beat, instrumental
Caption: Lo-fi hip hop beat, dusty vinyl texture, mellow chords, soft drums, warm bass, instrumental
Lyrics: [Instrumental]
This output tests whether a compact list of sonic materials can define an instrumental without lyrics competing for attention. The request gives a texture, harmony mood, drum character, bass quality, and an explicit instrumental instruction. Pick this approach for background beds, product cutdowns, or mood boards. It is less useful when a clip needs a memorable sung line, because the prompt intentionally removes that requirement.
3. Punk rock chorus
Caption: High energy punk rock anthem, fast drums, crunchy guitars, raw female vocals, big chorus
Lyrics: A short verse followed by a chant-ready chorus about owning the night.
This is the hardest vocal brief in the set because it asks for speed, distorted guitars, a raw vocal character, and a large chorus in only 10 seconds. The player is useful as a stress test for energy and attack. Keep lyric lines short when testing this kind of genre. Dense writing can consume the whole clip before the chorus has a chance to land.
4. Afrobeat dance groove
Caption: Afrobeat dance groove with bright percussion, bouncy bassline, syncopated guitars, uplifting female vocals
Lyrics: A verse and chorus built around movement, drums, and rhythm.
This brief separates groove ingredients clearly: percussion, bass movement, guitar rhythm, mood, and vocal type. That makes it a useful template for dance-oriented prompts. If the groove matters more than the lyric, shorten the text further and specify a BPM. If the vocal line matters more, retain the short chorus but avoid adding extra instruments that will compete for the same ten seconds.
5. Trailer tension, instrumental
Caption: Cinematic orchestral trailer cue, rising strings, big brass hits, deep drums, dramatic tension, instrumental
Lyrics: [Instrumental]
The trailer cue asks for an arc: rising strings, impact hits, and deep percussion. In a 10-second clip, that is enough to test an opening build or a transition sting, not a full trailer structure. Choose this model and prompt style for a fast cinematic sketch. For an edit point, state the desired hit or cutoff directly in the caption and test several seeds rather than expecting one run to solve timing.
6. Cozy indie-folk verse
Caption: Acoustic indie folk song, intimate fingerpicked guitar, soft percussion, airy female vocals, cozy mood
Lyrics: A compact verse and chorus about coffee, rain, and a quiet start to the day.
This is the least crowded vocal brief. It names one lead instrument, restrained rhythm, a vocal timbre, and a mood. That is a good choice when phrasing and intimacy matter more than impact. Use it for creator intros, reflective social clips, and song sketches. If the result needs a stronger refrain, put the chorus first and make its words shorter.
When to pick ACE-Step v1.5
Pick ACE-Step v1.5 on Wiro when a music idea needs a quick, controlled first pass across vocals or instrumentals, especially when the brief benefits from explicit duration, lyrics, genre, and BPM controls. The model documentation also supports reference audio, multiple vocal languages, and a planning option, so it offers more room to refine than a caption-only tool. Start with 10 seconds to test the sound palette. Move to 30 seconds only after the hook, groove, or texture is close.
For an instrumental, say so plainly and keep the caption focused on rhythm, texture, and the lead sound. For vocals, use a small number of short lines and label sections such as verse and chorus. For a fixed tempo, set BPM rather than relying on an adjective alone. A seed can help make comparisons repeatable, but the saved examples above did not record one.
Further reading and related tests
ACE-Step publishes its open-source implementation on GitHub and its model card on Hugging Face. Both sources were checked successfully before this update. For another short music test on this blog, read Google Lyria 3: 5 Prompt Tests for Short Music Clips. For music built around footage rather than a text-only brief, see Video Background Music v2: 4 styles for one clip and MMAudio: 4 Video-to-Audio Before/After Tests.
Try the same six prompt patterns in Text To Song on Wiro, then change one variable at a time: duration, BPM, lyric length, seed, or the instrumental switch.
