AI Voice Generator & Cloner — TTS + voice cloning in 646 languages

Turn text into natural speech, or clone a voice from a short sample, over a simple HTTP API. Design a voice by describing it (gender, age, pitch, accent, whisper) or clone one from ~10 seconds of reference audio, then make it say anything — in 646 languages, with expressive tags like [laughter], [sigh], [surprise-ah]. You get back a ready-to-use WAV file. No GPU, no model hosting, no subscription — pay per use.

• 🗣️ Text-to-speech — describe a voice and generate speech instantly
• 🧬 Voice cloning — clone from a reference sample (URL or base64) and speak in that voice
• 🌍 646 languages — Spanish, English, Japanese, Italian, Portuguese, and hundreds more
• 🎭 Expressive — inline tags for laughter, sighs, surprise, questions; adjustable speed & quality
• ⚡ HTTP API (Standby) + one-shot runs + usable from AI agents (MCP)
• 💵 Pay per request + per 1,000 characters — no monthly plan

Sample output

{
  "ok": true,
  "mode": "design",
  "url": "https://api.apify.com/v2/key-value-stores/…/records/voice_design_lq3k.wav",
  "contentType": "audio/wav",
  "durationMs": 3200,
  "characters": 61,
  "billedChars1k": 1,
  "language": "es"
}

The url points at a WAV 24 kHz mono file you can download or stream directly.

How to use

As an HTTP API (Standby — recommended)

Base URL: https://synthetic-ia--ai-voice-generator.apify.actor with header Authorization: Bearer <APIFY_TOKEN>.

Endpoint	Body	Does
POST /tts	{ text, language?, gender?, age?, pitch?, whisper?, accent?, speed?, quality? }	Design a voice and speak → WAV
POST /clone	{ text, referenceAudioUrl | referenceAudioBase64, referenceText?, rightsConfirmed: true, language?, speed?, quality? }	Clone a voice and speak → WAV
POST /generate	{ mode: "design" | "clone", … }	Unified endpoint
GET /	—	Returns this endpoint list as JSON

# Text-to-speech (design a warm female narrator, in Spanish)
curl -X POST "https://synthetic-ia--ai-voice-generator.apify.actor/tts" \
  -H "Authorization: Bearer $APIFY_TOKEN" -H "Content-Type: application/json" \
  -d '{"text":"Bienvenidos al episodio de hoy. [laughter]","language":"es","gender":"female","age":"young_adult","pitch":"moderate"}'

# Clone a voice from a reference sample and make it read a line
curl -X POST "https://synthetic-ia--ai-voice-generator.apify.actor/clone" \
  -H "Authorization: Bearer $APIFY_TOKEN" -H "Content-Type: application/json" \
  -d '{"text":"This is my cloned voice.","referenceAudioUrl":"https://example.com/me-10s.wav","rightsConfirmed":true}'

As a one-shot run

Set the input (mode, text, voice options) and run once. The result WAV is saved to the run's key-value store as OUTPUT, and a record with the URL is pushed to the dataset.

From an AI agent

The Actor is exposed over Apify's MCP integration, so an agent can call it as a "generate voice" / "clone voice" tool and get back an audio URL.

Input

The main fields (see the full input schema in the Console):

• mode — design (describe the voice) or clone (from a sample).
• text — what to say (max 5000 characters). Supports expressive tags.
• language — ISO code (es, en, ja…) or English name.
• Design: gender (female/male), age (child…elderly), pitch (very_low…very_high), whisper (ASMR), accent (English voices).
• Clone: referenceAudioUrl or referenceAudioBase64 (~10s+ clean speech, http(s), max 12 MB), optional referenceText, and rightsConfirmed: true (see Legal).
• quality — fast / balanced / high (default). speed — 0.5–2.0.

Pricing

Pay only for what you generate — no subscription:

What	Price (Free tier)
Per voice request	$0.05
Per 1,000 characters	$0.10
Voice cloning surcharge (per clone)	$0.15

Prices drop on higher usage tiers (down to $0.03 / $0.06 / $0.08). Platform usage is included.

Examples: a 200-character TTS line ≈ $0.15. A 1,000-character narration ≈ $0.15. A 500-character clone ≈ $0.30. That's typically 2–3× cheaper than subscription TTS services for text-to-speech, and cloning is pay-per-use with no monthly plan.

Need MP3 instead of WAV? Chain the ffmpeg API Actor to transcode /extract-audio or /convert.

Related Actors

• ffmpeg API — convert the WAV to MP3, or mix voice over video.
• Live Stream Clipper — clip live streams; pair with voiceover.

FAQ

Which languages? 646, including Spanish, English, Japanese, Italian, Portuguese, and many more. Pass the ISO code or English name.

How long can the text be? Up to 5,000 characters per call. Split longer scripts into multiple calls.

What audio do I get? WAV, 24 kHz, mono. Convert to MP3 with the ffmpeg API Actor if needed.

Why the first call is slow. The engine may cold-start (~40–60 s) if it has been idle; subsequent calls are fast. The per-request fee covers this.

Do you keep my audio or reference sample? The output WAV is stored in your run's key-value store. Reference samples are sent to the voice engine only to produce your clone.

Legal

Voice cloning requires consent. To use clone mode you must set rightsConfirmed: true, confirming you own the voice or are authorized to use it. Do not clone a person's voice without their permission, and do not use this Actor to impersonate, deceive, or infringe anyone's rights. You are responsible for the text you synthesize and the reference audio you provide. Misuse (fraud, harassment, non-consensual voice cloning, or violating a platform's terms) is prohibited.