How to Use ACE-Step 1.5 in AI FILMS Studio: Lyrics to Music Guide

ACE-Step 1.5 generates a complete track from a list of style tags, with lyrics as an optional extra. Leave the lyrics field empty and it returns an instrumental. Fill it in and it sings. That single behaviour makes it the most useful music model in the Studio for filmmakers, who need score and background beds far more often than they need songs with vocals.

It is the default model on the Lyric to Music generation type, it handles more than 50 languages, and it gives you direct control over track length up to four minutes.

Opening the Music Generator

Go to the AI FILMS Studio music workspace or click Create Music in the top navigation.

Setting the Generation Type

ACE-Step 1.5 lives under Lyric to Music. The Text to Music option offers a different model with a different set of controls, so the generation type has to be right before the model picker will show you ACE-Step 1.5.

Selecting the Model

Open Select Model. The picker is searchable and lists two options.

Writing the Style Tags

The field is labelled Style Tags for this model, and the label is doing real work. It expects comma separated descriptors rather than a written sentence.

Adding a tempo in beats per minute works well here. A tag list like "steampunk, electro swing, jazz, piano, ticking clock sounds, upbeat, male crooner, brass section, groovy, 110bpm" gives the model a genre, a texture, a vocal character and a speed, and the output holds together far better than a two tag list would.

Lyrics, and Leaving Them Out

The Lyrics field is marked optional, and the label states plainly that omitting it produces an instrumental. This is the control that makes ACE-Step 1.5 worth reaching for on a film project.

Eight buttons sit under the field. Intro, Verse, Chorus, Bridge and Outro insert a lowercase structure marker in square brackets at the cursor, such as [verse]. ACE-Step 1.5 reads those markers natively and uses them to segment the arrangement, not only the vocal.

The other three shape timing. New Line inserts a single line break. Pause inserts a double line break, which the model reads as a rest. Accompaniment inserts a ## marker, used in pairs to wrap a passage you want treated as an instrumental direction rather than sung words.

[verse]
Ticking tocking goes the clock
Turning keys and pick the lock

[chorus]
Steam is rising, gears go round
Listen to that brassy sound

Duration

The Duration slider is the control ACE-Step 1.5 has and the other music models do not.

Set duration to match the cut you are scoring rather than generating long and trimming. The model paces the arrangement across whatever length you give it, so a 30 second setting produces a piece that resolves at 30 seconds instead of a four minute track cut short.

Seed

This is the fastest way to refine a cue. Generate with a random seed until something is close, note the seed number, then iterate on the tags with that seed locked.

Parameter Reference

Control	Values	Default	What it changes
Select Generation Type	Text to Music, Lyric to Music	Text to Music	Which models are available. ACE-Step 1.5 needs Lyric to Music
Select Model	ACE-Step 1.5, MiniMax Music 2.5	ACE-Step 1.5	Which controls appear below
Style Tags	Up to 3,000 characters, comma separated	Empty	Genre, mood, instruments, tempo
Lyrics	Up to 3,000 characters, optional	Empty	Words sung. Empty produces an instrumental
Duration	1 to 240 seconds	60 seconds	Length of the generated track
Seed	Any integer, or -1	-1	Reproducibility. -1 gives a random result each run

Cost scales with duration at 0.30 credits per second, so a 60 second cue costs 18 credits and a full 240 second track costs 72. The figure appears in the form before you submit and updates as you move the slider.

Running It in the Nodes Graph Editor

The Nodes Graph Editor exposes ACE-Step 1.5 through a node named Text To Music. That name refers to the node family. The model selector inside it still offers ACE-Step 1.5.

The node view offers a different set of style tag buttons to the workspace, covering cinematic, orchestral, epic, emotional, ambient, electronic, rock, pop, jazz, classical, hip hop, EDM, country, folk, metal, punk, reggae, blues, soul and funk. The cinematic and orchestral tags in particular are more useful for scoring than the workspace list.

Use the node route when the track is one step in a longer chain. Generating a cue and passing it into a video or lipsync node keeps the whole sequence in one graph.

Limitations

No stems. Output is one mixed file. You cannot rebalance the vocal against the instrumentation afterwards, so plan the mix through the style tags.

Duration is a target, not a hard cut. The model paces an arrangement to the length you set. Very short durations under about 10 seconds sometimes end before the phrase resolves.

Structure markers only work in the lyrics field. Putting [chorus] in the Style Tags field does nothing. The markers belong with the words.

Four minutes is the ceiling. For anything longer, generate sections separately with a fixed seed and matching tags, then assemble them in the editor.

Tips for Better Results

Add a bpm tag. Naming a tempo such as "110bpm" gives the model a concrete target and makes successive generations easier to cut together.

Leave lyrics empty first. For scoring work, get the instrumental right before deciding whether the cue needs a vocal at all. Most film cues do not.

Lock the seed once you are close. Random seeds are for exploring. A fixed seed is for refining, because it isolates the effect of each tag you change.

Match duration to the cut. Generating at the length you actually need produces a piece that resolves properly, which saves you fading out an unfinished phrase.

ACE-Step 1.5 is also released openly, and it runs locally under an MIT license and scores above Suno v5 on SongEval benchmarks on consumer hardware with under 4GB of VRAM. For a model on the same generation type that requires lyrics and returns higher audio settings, the MiniMax Music 2.5 lyrics to music workflow covers sample rate, bitrate and output format. For style driven songs without writing lyrics, the Suno text to music workflow in AI FILMS Studio covers the Text to Music generation type.

---

Sources

ACE-Step | AI FILMS Studio | SongEval