Now you can make AI music OFFLINE!!! Free unlimited AI music generator

Key Concepts

• Open-source AI music generation: Using freely available AI models to create music.
• Text-to-music generation: Creating music from text prompts describing the desired genre, style, and lyrics.
• Meta tags: Keywords within lyrics (e.g., verse, chorus, drop) that guide the AI's song structure.
• Guidance scale: A setting that controls how closely the AI follows the text prompt and lyrics.
• Audio-to-audio: Generating music in the style of a reference audio track.
• Impainting: Editing specific sections of a generated song.
• Retake: Regenerating a song with slight or significant variations.
• Virtual environment: An isolated environment for running Python code with specific dependencies.

Acep: The Best Open-Source AI Music Generator

This video introduces Acep, a new open-source AI music generator developed by Ace Studio and Stepf fun. It is presented as a superior alternative to other open-source options like Yu, and comparable to commercial tools like Udio, Suno, and Refusion. The video covers demos, usage instructions, tips and tricks, and local installation.

Demos and Capabilities

Acep allows users to generate music by specifying a genre and optional lyrics. The video showcases several demos:

• General Demo: Demonstrates the tool's ability to create a song with lyrics and a defined structure.
• Alternative Rock: Shows Acep's ability to create a realistic and dynamic rock song with vocals and instrumentals.
• Electronic Rap: Highlights the tool's ability to generate electronic rap music, noting occasional mispronunciations that can be corrected through regeneration or impainting.
• Dubstep: Demonstrates the tool's ability to create intense dubstep music with heavy drops and echoing growls.
• Ethereal Dark Liquid Deep Baseline: Showcases the generation of a song with a specific mood and vocal style.
• Instrumentals: Demonstrates the generation of instrumental pieces in genres like saxophone jazz, sonata, tango, and psychedelic trance.
• Multilingual Support: Showcases music generation in Chinese, French, German, and Japanese (anime/J-pop). The speaker admits to not knowing if the German lyrics are correct.

Using Acep

Acep can be used online via a Hugging Face demo or installed locally. The interface includes settings for:

• Audio Duration: Specifies the length of the generated song. Setting it to -1 results in a random duration between 30 seconds and 4 minutes.
• Tags: Defines the genre or desired sound of the music.
• Lyrics: Allows users to input lyrics for the song.
• Infer Steps: Controls the number of iterations the AI goes through during generation. The default value of 27 is suggested as a sweet spot.
• Guidance Scale: Determines how closely the AI follows the prompt and lyrics. Separate guidance scales can be set for the text prompt and lyrics.
• Euler: Selects the algorithm used for music generation.

Tips and Tricks for Better Generations

The video provides several tips for improving the quality of Acep generations:

• Meta Tags: Using tags like "verse," "pre-chorus," "chorus," "drop," "intro," "outro," and "bridge" helps structure the song.
• Echo Effect: Adding words in parentheses at the end of a line can create an echo effect.
• Abbreviations: Spacing out letters in abbreviations (e.g., "R T X G P U") ensures they are pronounced correctly.
• Numbers: Typing out numbers (e.g., "fifty ninety" instead of "5090") improves pronunciation.
• Genre Specificity: Focusing on specifying the genre, instruments, or mood in the tags, rather than trying to emulate specific artists or BPM.
• Instrumental Generation: Using "[instrumental]" or "inst" in the lyrics field generates a completely instrumental song.

Audio-to-Audio Feature

The audio-to-audio feature allows users to upload a reference track and generate music in its style. The "influence" slider controls how closely the generated music resembles the reference. A value between 0.2 and 0.3 is recommended.

Editing and Extending Songs

Acep offers several editing features:

• Retake: Regenerates a song with variations controlled by a "variance" slider.
• Repaint: Edits specific sections of a song by specifying the start and end times.
• Edit: Allows users to change the lyrics of a song while keeping the same instrumental and vocal style.
• Extend: Extends a song by a specified duration to the left or right.

Local Installation

The video provides a step-by-step guide to installing Acep locally:

1. Install Git: Download and install Git from the official website.
2. Clone the Repository: Use git clone to download the Acep repository from GitHub.
3. Install Miniconda: Download and install Miniconda (a minimal version of Anaconda) from the official website.
4. Add Anaconda to Path: Add the Anaconda scripts directory to the system's PATH environment variable.
5. Create a Virtual Environment: Use conda create --name ace_step python=3.10 to create a new virtual environment.
6. Activate the Environment: Use conda activate ace_step to activate the virtual environment.
7. Install Dependencies: Use pip install torch torchaudio --index-url https://download.pytorch.org/whl/cu118 (for Windows with NVIDIA GPU) to install PyTorch and other dependencies.
8. Install Acep: Use pip install -U astep to install Acep and its core dependencies.
9. Run Acep: Navigate to the Acep directory in the command prompt, activate the virtual environment, and run acep --port 7865.

The minimum VRAM requirement is 8 GB.

Conclusion

Acep is presented as a powerful, flexible, and free open-source AI music generator with capabilities for text-to-music generation, audio-to-audio transfer, and various editing options. The video encourages viewers to experiment with the tool and share their creations and troubleshooting tips in the comments. The speaker also mentions a giveaway for a Dell Precision 5690 laptop with an RTX 5000 ADA GPU.