Google Cloud Text-to-Speech

Key advantages

• Noticeably better voice quality than edge-tts, especially on Neural2 and Studio voices
• A permanent free tier (not a time-limited trial), comfortably enough for regular editorial use
• Official, documented REST API, easy to automate inside an existing script

Who it's for, and why

Content creators who want noticeably better voice quality than a free engine like edge-tts, and who accept creating a Google Cloud project with active billing in exchange.

• Audio narration of blog articles with a Neural2 or Studio voice
• Voice generation for voice interfaces or assistants
• Quick voiceover prototyping before choosing a higher-end tool (Studio, ElevenLabs)

Getting started

1. 1Create or pick a project at console.cloud.google.com and enable billing (a card is required even to stay within the free tier).
2. 2Enable "Cloud Text-to-Speech API" under APIs & Services → Library.
3. 3Create an API key under APIs & Services → Credentials, and restrict it to this API only for safety.
4. 4Call the REST API: `POST https://texttospeech.googleapis.com/v1/text:synthesize?key=YOUR_KEY` with a JSON body `{"input":{"text":"..."},"voice":{"languageCode":"fr-FR","name":"fr-FR-Neural2-G"},"audioConfig":{"audioEncoding":"MP3"}}` (audio comes back base64-encoded in the response).

Things to watch out for

• Requires an active Google Cloud billing account (a card on file), even to stay within the free tier: unlike edge-tts, this is no longer "zero risk" without budget monitoring.
• The free tier caps at 1 million characters/month for Neural2 and WaveNet voices (4 million for Standard voices, less natural-sounding); usage beyond that is billed per character.
• A Google Cloud budget alert must be configured manually (Billing → Budgets & alerts); it isn't enabled by default when the project is created.
• The `text:synthesize` request is capped at roughly 5,000 bytes of text: a full article has to be split into several calls and the resulting audio concatenated.