Make a finished track with SteppeMind Composer

Composer or Maestro?

Composer audio cannot be opened as the original note score. Stem separation gives you audio parts; it does not turn them into MIDI either.

Choose a Composer model

Aurora is designed for a versatile premium result, Aurora 2 for a new generation of finished songs with expressive vocals and detailed arrangements, Pulse for quick previews, Prism for closer prompt adherence, Forge for vocal-aware and part operations, and Horizon for long instrumental music, continuation and targeted replacement. The workflow stays consistent: describe the task, audition the result and only then apply it.

For Aurora 2, describe the mood, instruments, desired duration and song structure in your prompt. If you want vocals, include the language and lyrics; for an instrumental composition, explicitly request “no vocals.” Duration is treated as a preference, so the finished track may be a different length. Check the selected model’s panel for current limits and cost.

Horizon accepts a negative prompt and adjustable change strength.

Forge can also change style, add or complete music, create variations and work with individual parts. Read Models and capabilities for a full comparison, current limits and the safe way to transform source audio.

Write a useful description

Begin with the purpose, mood and development of the music. Then add the main instruments, the character of the rhythm and whether you want vocals. Include a tempo if it matters to the project. A long list of conflicting requests makes the result harder to judge.

For example: “An energetic instrumental track for a night drive. Strong bass, tight drums and a dark melody. A restrained introduction followed by a fuller main section. No vocals.”

“Leave only piano and atmosphere in the first section” is more useful than “make it high quality”: you can hear whether the first request was followed.

Start a generation

1. Open the intended project and the SteppeMind AI agent in the top bar.
2. Select SteppeMind Composer in the conversation input, then choose a model such as Aurora 2.
3. Check availability and credit information. Send your request once and watch its task status.
4. Follow the progress in the agent window. Avoid sending identical requests just because the result is not immediate; processing takes time.
5. When the audio card appears, use its play button. Listen to the beginning, a main section and the ending.

Bring a good result into the project

The result card offers actions to apply the audio to the project, repeat the task, discard the result or separate it into stems. Preview first, then apply.

All Composer models support BPM in the result card. When the model does not supply a tempo, the app automatically analyzes the source audio and marks the estimate with ≈. BPM is saved with the result and survives a page reload.

Use ÷2 or ×2 to correct a half-time or double-time estimate, or enter a value manually. If the source is temporarily unavailable, choose Retry BPM. Music without a steady rhythm may not have a reliable estimate; you can leave the field empty.

Applying a result automatically sets BPM in an empty project. Existing audio or MIDI arrangements keep their tempo. A manual correction takes priority over automatic analysis.

After applying, check the clip in Studio, its start position and its level against other tracks. Save the project and wait for confirmation. To keep a separate audio file, use the download action in the result card or export the finished mix.

Change one part instead of the whole track

Try stem separation if you want quieter drums, no voice or just the bass. If you need to change individual notes, Maestro is a better starting point. Repeating a Composer generation may change the entire musical result, not just one instrument.

If a task fails, read the notification, check your connection and credit balance, then retry once. For repeated failures, keep the error message and contact support through Feedback. Never send passwords or access keys.