Fully Local Generation
Runs Piper, RHVoice, and eSpeak NG on the AI-tools server without an external TTS API.
Fully local speech across a broad language catalog
Convert multilingual text into MP3 speech with local Piper, RHVoice, and eSpeak NG engines—without sending text to an external AI provider.
/api/text-to-speech-api/v1/generate-speech
What it does
Use GET /voices to discover every language and voice currently installed on the server. Then POST /generate-speech with text, a BCP 47 language code such as en-US, hr-HR, or mk-MK, and optionally a voice ID returned by the catalog. If voice is omitted, the service chooses the highest-priority local voice for that language. English defaults to piper:en-us-lessac-medium, Croatian to rhvoice:karmela, and Macedonian to rhvoice:suze. Other installed RHVoice voices are preferred before the broad eSpeak NG fallback. Audio is generated on this server, converted to MP3, and deleted 3,600 seconds after creation. No submitted text or audio is sent to OpenRouter or another external TTS provider.
Runs Piper, RHVoice, and eSpeak NG on the AI-tools server without an external TTS API.
Exposes the currently installed languages, engines, voice IDs, quality types, and defaults.
Uses Piper Lessac for English, RHVoice Karmela for Croatian, and RHVoice Suze for Macedonian by default.
Returns a non-guessable HTTPS MP3 URL that becomes unavailable and is deleted after one hour.
Stable endpoints
/api/text-to-speech-api/v1/generate-speechGenerate an MP3 locally using the selected language and optional voice ID.
/api/text-to-speech-api/v1/voicesReturn the live local catalog grouped by language, including voice IDs, engines, quality types, and the default voice for each language.
Simple monthly plans
Plus includes 10× the Free quota. Pro includes 500× the Free quota.
Free access for testing and light usage.
10× the Free request quota.
500× the Free request quota for production workloads.