SongGeneration

SongGeneration (Tencent)

End-to-end music and song generation from text prompts.

• Music-first model. Designed specifically for generating full songs, including melody, instrumentation, and vocal structure, rather than stitching together short audio clips.

• Text-conditioned control. Uses natural language prompts to guide genre, mood, style, and overall musical direction.

• Structured outputs. Produces coherent musical pieces with consistent rhythm and progression, not just loops or sound effects.

• Heavy but expressive. Requires significant compute compared to simple audio models, trading speed for richer musical structure.

• Research-focused release. Intended for experimentation and evaluation, not lightweight or real-time music generation.

Why pick it for Norman AI?

SongGeneration is a good fit for exploring text-to-music and song synthesis. It’s useful for testing creative audio workflows, prompt control over music style, and comparing generative music models beyond basic audio generation.

response = await norman.invoke(
    {
        "model_name": "songgeneration",
        "inputs": [
            {
                "display_title": "Prompt",
                "data": "Create a groovy, rhythmic remix of the input audio."
            },
            {
                "display_title": "Prompt",
                "data": "/Users/alice/Desktop/sample_input.aac"
            }
        ]
    }
)