sanjay3290/ai-skills

Overview

This skill converts text and documents into natural-sounding audio using Google Cloud Text-to-Speech. It supports Neural2, WaveNet, Studio, and Standard voices across 40+ languages and outputs MP3, WAV, or OGG files. The tool handles single-voice narration, multi-speaker podcast scripts, and automatic chunking for large inputs.

How this skill works

Provide text directly or supply a document (TXT, MD, PDF, DOCX). The skill reads your input, optionally extracts text from files, chunks long content at sentence boundaries, and calls Google TTS with chosen voice, rate, pitch, and encoding. For podcasts, supply a JSON script of alternating speakers and the tool synthesizes each speaker with a different voice and merges the audio into one file.

When to use it

• You want to narrate a document, article, or report for listening.
• You need to convert text to a downloadable MP3/WAV/OGG audio file.
• You want to generate a short podcast or multi-voice recording from a script.
• You need batch or automated TTS for documentation or training material.
• You want higher-quality voices (Neural2/WaveNet) across many languages.

Best practices

• Set GOOGLE_TTS_API_KEY via env var or config JSON before running.
• Use Neural2 voices for best quality/cost balance; default voices provided.
• Preprocess long documents into clean prose and let the script auto-chunk.
• Choose a speaking rate ~0.85–0.95 for technical content and keep pitch modest.
• Keep individual turns under 4000 chars when creating multi-speaker scripts.

Example use cases

• Narrate a PDF user manual into an MP3 for accessibility or offline listening.
• Create a two-host podcast from a document by converting it into a conversational JSON script.
• Generate voiceovers for training videos or e-learning modules using specific voices and rates.
• Produce audio versions of blog posts or long-form articles for a listening audience.
• Automate daily briefing recordings by feeding generated summaries into the TTS script.