JASCO

Preview：

Introduce：

JASCO is a text-to-music generation model that combines symbolic and audio-based conditions to generate high-quality music samples based on global text descriptions and fine-grained local control. JASCO is based on the flow matching modeling paradigm and a novel conditional approach that allows music generation to be controlled both locally (e.g. chords) and globally (textual descriptions). Information related to specific controls is extracted through an information bottleneck layer and temporal ambiguity, allowing the combination of symbolic and audio-based conditions in the same text-to-music model.

Stakeholders:
JASCO is for music creators, music theorists, and anyone interested in music generation technology. It can help users generate music that fits specific styles and emotions through text descriptions, providing new tools and sources of inspiration for music creation.
Usage Scenario Examples:

• Music creators use JASCO to generate music with a specific style based on text descriptions.
• Music theorists use JASCO to explore the effects of different textual descriptions on music generation.
• Educators use JASCO as a teaching tool to help students understand the relationship between music and text.

The features of the tool:

• Supports global text description and fine-grained local control.
• Based on flow matching modeling paradigm and novel conditional method.
• The technology of information bottleneck layer and time ambiguity is applied.
• Can combine symbolic and audio based conditions.
• Generation quality and condition compliance were assessed using objective indicators and human studies.
• Compared to the baseline model, it is comparable in generation quality while providing more flexible control.

Steps for Use:

• Visit JASCO’s official website.
• Learn the fundamentals and functionality of JASCO.
• Select or enter a text description of the music you want to generate.
• Select local control conditions, such as chords or melodies, as needed.
• Adjust other build parameters, such as rhythm or style.
• Start the music generation process and wait for the results.
• Evaluate the generated music samples and make adjustments based on feedback.

Tool’s Tabs: Music generation, text to music