Kling Audio Clone

Problem It Solves

When you need custom speech in a specific voice, generic TTS often produces a neutral sound rather than consistently restoring the timbre, cadence, and speaking style of a reference recording. This skill targets voiceover, character speech, and batch text-to-speech workflows by turning “upload reference audio + provide prompt or parameters” into an executable, pollable generation task that can fit into existing content or automation pipelines.

How It Works and Key Capabilities

• The skill is a thin client over the dLazy hosted API; the core invocation is dlazy kling-audio-clone.
• Every request requires a dLazy API key, which can be configured through the CLI or supplied via the DLAZY_API_KEY environment variable.
• When local audio paths are provided, the files are uploaded to dLazy media storage so the model can read them; generated outputs are typically hosted on files.dlazy.com.
• It supports async execution: passing --no-wait returns a generateId and task status, which can be polled with dlazy status --wait.
• Error handling covers common pipeline failures, including unauthorized, missing required parameters, local file read errors, insufficient balance, server-side errors, and async task failures caused by safety policies.

Boundaries and Cautions

• Generation depends on a dLazy account, API key, balance, and service terms; it is not a fully local inference solution.
• The API key is scoped to the dLazy organization and can be rotated or revoked from the dashboard.
• If the response returns insufficient_balance, the workflow should surface a balance issue; if it returns unauthorized, the API key must be configured before retrying.
• Final speech quality still depends on reference audio quality, model behavior, and prompt constraints, so it should not be assumed to reproduce the original speaker without limits.