AI Song Generator

The Problem It Solves

For developers needing to rapidly prototype or create custom audio content, producing a complete song from scratch involves complex steps like lyric writing, composition, arrangement, and recording, which typically require specialized software and audio engineering expertise. This process is time-consuming and has a high barrier to entry. This skill encapsulates this workflow via an AI model into a programmatically callable command-line tool, allowing developers to generate complete, audition-ready song audio and project files directly from textual descriptions (e.g., theme, mood, style, lyrics), drastically shortening the path from创意 to deliverable.

Core Workflow & Capabilities

The core of this skill is the aimv command-line tool, which serves as the unified interface to interact with the backend AI generation service and the 海绵音乐 (Haimian Music) platform. The workflow revolves around key steps:

1. Authentication: First-time use requires linking with a 海绵音乐 account via the aimv init command. The tool generates a dedicated connection page URL; after user login, an API Key is obtained to complete authentication. All subsequent operations use this API Key for identity verification.
2. Song Generation: This is the core capability. Developers invoke the aimv song create command and provide a text prompt (e.g., --prompt "A light pop song about a summer trip"). The AI model generates complete song audio (typically two candidate versions), lyrics, and metadata based on the prompt. The generation process occurs asynchronously in the cloud.
3. Status Query & Management: Generation tasks may take time. Their status and results can be checked via commands like aimv session list or aimv song status. All generated works are synced to the 海绵音乐 platform, and can be further played, managed, and downloaded via the accompanying WeChat Mini Program or the lexuan.club web interface.
4. Error Handling: The tool provides structured error codes. For instance, a code=auth_required error returns a login URL; code=insufficient_points returns a payment URL. This allows an Agent to automatically parse errors and guide users to the next step (e.g., login or payment) instead of merely reporting an error.

Boundaries & Considerations

• Environmental Dependencies: Skill execution relies on network connectivity. In some restricted Agent sandbox environments, if the aimv command fails due to network limits, try setting the environment variable AIMV_FORCE_CURL=1 to make the tool use the system curl for transport. If aimv is completely unusable, fallback to directly calling its REST API with curl.
• Pay Model & Quotas: Song generation consumes credits or membership quota on the 海绵音乐 platform. When encountering errors like code=insufficient_points (402) or code=quota_limited (429), the user must be directed to visit the returned pay_url to top up credits or subscribe to a membership plan. This is a user-participatory payment step.
• Content Dependence: Generation quality heavily depends on the AI model's capabilities and the quality of the user's text prompt. Copyright, originality, and style of generated songs may be ambiguous, and the level of finish may not meet all professional demands. It is best suited for scenarios like rapid exploration, demo creation, or background music for short videos.