AI Pop Audio Synthesizer
This skill handles the synthesis and verification of music audio assets. It supports both local workspace execution and remote/managed agent execution via GCS staging.
Instructions
-
Resolve Output Storage Mode:
- Local Mode (Default): Pass
local_pathpointing to the target local workspace and set naming scheme (track[N].wav). - Managed Agent / GCS Mode: When executing inside a managed agent container where the MCP server runs remotely on Cloud Run, pass
output_gcs_bucket(or setGENMEDIA_BUCKET) andoutput_filename(track[N].wav). The generated file will be staged directly into the mounted GCS environment bucket.
- Local Mode (Default): Pass
-
Invoke Lyria 3 Pro:
- Call
lyria_generate_musicwith the target prompt using the resolved storage parameters. - Robust Model Fallback: If
lyria-3-pro-previewencounters API response issues or rate limits, immediately fallback and retry usinglyria-3-clip-previeworlyria-002to ensure consistent and uninterrupted synthesis.
- Call
-
Verify Stream Container:
- Run the shell
filecommand (or inspect GCS metadata) to check the underlying container encoding (e.g.file track1.wav). - CRITICAL: If the tool reports that the file contains an MPEG layer III (MP3) stream, rename the file extension to
.mp3(e.g.track1.mp3) instead of saving it as a.wav. This ensures browser decoding compatibility.
- Run the shell
-
Format Check:
- Verify the file exists, has a non-zero size, and matches the configuration expected by the Go compiler.