AI-Powered Personalized Song Maker Skill

Solving a Specific Problem

Crafting a song tailored to a personal story for a special occasion traditionally involves multiple stages like lyric writing, composing, and arrangement, often requiring professional help at high cost and with unpredictable timelines. This skill automates the end-to-end process from story to song, enabling users to quickly generate customized, audible music by providing structured inputs, ideal for emotionally-driven scenarios like birthdays, weddings, or anniversaries.

How It Works: Core Capabilities and Key Steps

The skill implements a generative music creation pipeline with the following core capabilities and steps:

• Input Collection and Story Structuring: The skill first collects a set of mandatory inputs, including the occasion, recipient, real story details to incorporate, required names or phrases, target language, emotional direction, and privacy omissions. It then structures this into a story map encompassing protagonist, relationship, turning points, core message, emotional arc, etc., laying the groundwork for lyric creation.

• Free Creative Phase and User Confirmation: Before triggering any paid API calls, the skill performs free creative work. It generates a song title, a complete lyric draft with structure, and a production card describing genre, atmosphere, instrumentation, tempo, and vocal direction. Users can iteratively review and modify lyrics and direction at this stage at no cost until explicit confirmation is given.

• API Call and Music Generation: After confirmation, the skill queries the beatra.models.list endpoint to check available music generation models, character limits, and billing. It then uses the bundled client scripts/mcp_client.py as an MCP client to submit one precise generation request to beatra.music.generate. This request includes the frozen title, lyrics, production prompts, and model parameters; one call constitutes one paid boundary.

• Task Polling and Result Delivery: After submission, the skill obtains a task ID and polls via beatra.tasks.get until the task reaches a terminal state (success or failure). It delivers returned audio files or links and reports parsed model, segment count, duration, and usage. The skill also performs result review, checking for story accuracy, name pronunciation, chorus effectiveness, etc., and clearly communicates to the user what the AI can or cannot fully guarantee.

Applicability and Considerations

• Focused on the Full Workflow: This skill is dedicated to the "story -> lyrics -> song" core route. If input is finished lyrics needing composition, use the "lyrics-to-music" workflow; for re-recording an existing song, use the "cover song" workflow; to create visuals for music, route to the "music short" workflow.

• Clear Payment Model: Lyric writing, title drafting, and direction adjustments are free creative preparation. The sole point of payment occurs when the user confirms and calls beatra.music.generate to produce the song. Changing any confirmed core element (like title, lyrics, or model) is treated as a new logical payment request.

• Strict State and Transaction Management: The skill relies on task IDs and phase IDs for state tracking. For lost requests or failed tasks, it provides recovery logic based on beatra.tasks.list and beatra.tasks.get, but strictly prohibits resubmitting a generation request without confirming the terminal state. Task states like "queued" and "running" are normal progress, not failures.