day253-volcengine-ai-audio-tts_skill

Overview

This skill provides text-to-speech generation using Volcengine (ByteDance) speech services. It wraps the online one-shot HTTP TTS API (openspeech.bytedance.com) and saves audio files, request payloads, and response metadata for verification. Use it to produce narration, multi-language output, or to troubleshoot TTS parameters and credentials.

How this skill works

The skill sends a single HTTP request to Volcengine's TTS endpoint with a JSON payload that includes text, voice_type, encoding, sample rate, and prosody controls. It validates environment variables for app_id, token, and cluster, generates a unique reqid per request, stores the API response and the produced audio locally, and returns a normalized tts.generate response (audio_path, sample_rate, format, duration_ms, code). The script uses Python requests and saves logs and evidence under a configurable output directory.

When to use it

• You need narration or voice output for apps, videos, or accessibility features.
• You want multi-language TTS with selectable voices and adjustable speed/volume/pitch.
• You need a reproducible test to validate credentials, cluster, and voice configuration.
• You must archive requests, responses, and generated audio for compliance or QA.
• You are debugging 429 rate limits or tuning chunking for long texts.

Best practices

• Set VOLCENGINE_TTS_APP_ID, VOLCENGINE_TTS_TOKEN, and VOLCENGINE_TTS_CLUSTER in the environment or a .env file before running.
• Keep text chunks under the recommended byte limit (≤1024 bytes) or split long content and synthesize in segments.
• Use a UUID-based reqid for each request to ensure traceability and idempotency.
• Start with a minimal verification request to confirm credentials and voice/cluster mapping.
• Respect rate limits: back off or reduce concurrency on 429 responses and log retries.

Example use cases

• Generate a single MP3 narration for a short onboarding video using voice_type BV700_streaming.
• Produce multi-language prompts for an IVR system by iterating language and voice_type parameters.
• Validate a new Volcengine token and cluster by running a minimal test request and inspecting saved metadata.
• Create a dataset of audio samples by batching short texts and storing audio_path and response details.
• Adjust speed_ratio, pitch_ratio, and volume_ratio to tune a synthesized voice for accessibility needs.