Create your own AI-generated music with ACE-Step

ACE-Step, developed by ACE Studio and StepFun, is an open-source model that generates music from the lyrics and style tags you provide, allowing you to create songs, vocals, and instrumentals from simple text inputs. With its built-in tools, you can also refine your tracks by adjusting or regenerating specific parts without starting over.

This guide shows you how to install ACE-Step on Olares, generate your first track, explore different musical styles, and enhance your audio using the editing features built into the app.

Learning objectives

By the end of this tutorial, you will learn how to:

• Install ACE-Step on your Olares device.
• Generate songs with lyrics, tags, and stylistic controls.
• Locate and download your generated audio files.
• Refine tracks by adjusting the style, editing sections, extending the song, or using a reference clip to reshape it.

Prerequisites

Before you begin, make sure:

• Olares running on a machine equipped with an NVIDIA GPU.

Install and set up ACE-Step

With your Olares device ready, follow these steps to install ACE-Step and begin generating music.

Install ACE-Step

Follow these steps to install ACE-Step.

1. Open the Market app in your Olares web interface.
2. Use the search bar and type "ACE-Step".
3. Click Get, then click Install.
   
4. Wait a few minutes for the installation to complete.

Download required models on first-time launch

Once installation finishes, open ACE-Step from Launchpad.

Olares will automatically download and install required models. A Download Manager window will appear, showing model size and download progress.

After the download completes, the ACE-Step generation interface will open automatically.

Generate your first track

Follow these steps to set your parameters and begin music generation.

Set basic parameters

• Audio Duration: Drag the slider to choose the track length (up to 240 seconds). If left at the default (-1), the audio length will be random.

• Format: Select the audio format from MP3, ogg, wav, and flac.

  MP3 recommended

  It's recommended to change the default output format to MP3. This will result in much smaller file sizes, faster loading, and a better user experience.

• Lora Name or Path: Select a LoRA model if available. Currently, only a Chinese rap LoRA is supported.

• Tags: Enter descriptors for style, mood, rhythm, or instruments, separated by commas. For example:

• Chinese Rap, J-Pop, Anime, kawaii future bass, Female vocals, EDM, Super Fast`

• Lyrics: Enter your lyrics, ensuring you use structural tags for optimal organization and flow:

  • [verse] for the main verse part
  • [chorus] for the chorus part
  • [bridge] for the bridge

  Generate an instrumental-only track

  Enter the tag [instrumental] or [inst] in the Lyrics area.

Start generation

1. Click Generate when all parameters are set.
2. Once generation is complete, click the Play button to preview your track.

Save the generated music

You can save your generated music via two methods:

• Direct download: Click the download button in the upper right corner to save the audio file directly to your local device.

• From Olares Files:

  1. Open Files.
  2. Go to the following path: /Home/AI/output/acestepv2.
  3. Right-click the generated audio file and save it to your local device.

Optimize your audio

ACE-Step offers powerful tools to refine and modify specific parts of your generated track.

Regenerate the entire segment

You can generate a new version of the entire track.

1. Click the retake tab.
2. Adjust the variance slider to control how different the new version will be. The higher the value, the more different the song will be.
3. Click Retake and wait for the generation.
4. Click the Play button below to preview the style change.

Regenerate a specific section

You can update only a selected time range while keeping the rest of the track unchanged.

1. Click the repainting tab.
2. Adjust the Variance slider to control the degree of change in the new generation. The higher the value, the more different the song will be.
3. Adjust the slider under Repaint Start Time and Repaint End Time to set the period for the section you want to regenerate.
4. Select the source for repainting:
   • text2music: The original song generated via Text2Music.
   • last_repaint: The previous repainted version.
   • upload: The audio you uploaded.
5. Click Repaint and wait for the generation.
6. Click the Play button below to preview the result.

Edit lyrics

You can edit lyrics to modify specific lines without affecting the rest of the track.

1. Click the edit tab.
2. Copy the original lyrics and paste them into the Edit Lyrics area.
3. Modify only the specific lines of the lyrics you wish to change.
4. Under Edit Type, select only_lyrics.
5. Click Edit and wait for the generation.
6. Click the Play button below to preview the change.

Edit tags

You can edit tags to reset the style or timbre of the track.

1. Click the edit tab.
2. Enter the new style or timbre tags (e.g., hard rock or male tenor vocals) in the Edit Tags area.
3. In Edit Type, select remix.
4. Click Edit and wait for the generation.
5. Click the Play button below to preview the change.

Extend the audio

You can extend the length of the original track by adding new audio before or after it.

1. Click the extend tab.
2. Adjust the slider under Left Extend Length to add new generation before the original audio.
3. Adjust the slider under Right Extend Length to add new generation after the original audio.
4. Select the source to extend:
   • text2music: The original song generated via Text2Music.
   • last_extend: The previous extended version.
   • upload: The audio you uploaded.
5. Click Extend and wait for the generation.
6. Click the Play button below to preview the change.

Audio2Audio

You can create a new track based on a reference audio clip you upload. It analyzes characteristics such as timbre, rhythm, and style to produce a track with a similar feel.

1. Check the box to Enable Audio2Audio.
2. Upload an existing music clip to serve as the reference.
3. Adjust the Refer audio strength slider. A higher value results in music more similar to the reference track.
4. Select a Preset style, or keep the default.
5. Set other parameters as needed.
6. Click Generate to create new music with an atmosphere similar to the reference audio.