Voice Cloning & Verification

Not available on the public API

Voice cloning, emotion-vector TTS, tag-based expressive TTS, voice comparison, deepfake detection, and voice forensics are not available through the public FOTOhub API. Every endpoint earlier revisions of this page documented returns 404 in production:

• POST /v1/ai/voice/clone
• POST /v1/ai/voice/clone/preview
• POST /v1/ai/voice/clone/list
• DELETE /v1/ai/voice/clone/{voice_uid}
• POST /v1/ai/voice/emotional
• POST /v1/ai/voice/compare
• POST /v1/ai/voice/deepfake-detect
• POST /v1/ai/voice/forensics

There is no client.voice.clone(), client.voice.compare(), client.voice.deepfake_detect(), or client.voice.forensics() in the Python or TypeScript SDKs, and no voice_uid / emotion_vec parameter accepted anywhere on the public API.

Voice cloning is a real, priced capability in FOTOhub's internal model catalogue — GET /v1/models lists a billable model voice-clone ($0.797/request, is_active: true), and it runs on FOTOhub's self-hosted IDA Voice engine. But that model has no working caller today, on any surface:

• REST: not reachable through /v1/ai/generate/speech or any other documented endpoint. Being priced and active in the catalogue means it is billable once something calls it, not that a route exists today.
• MCP: the MCP server registers a voice_clone tool, but its implementation is a stub — it makes no backend call and always returns "Voice cloning is not yet available via MCP." The tool used to call POST /voice_clone/train on chatterbox-server, a route that doesn't exist and 404s; that call was removed, and the real cloning code (music-server's voice_clone.py / tts.py) still has nothing public in front of it. A tool being registered is not proof it does anything — see MCP Integration for what MCP tools actually work today. The same is true of the separate_stems MCP tool: registered, but a stub that always returns "Stem separation is not yet available via MCP" rather than calling music-server's real (but unbilled, so not public-facing) /stems route.
• Dashboard: the MCP stub's own error message points users at "the FOTOhub dashboard's Voice Cloning page" instead — that first-party UI is the one place this actually runs today, outside the scope of this developer API reference.

The public POST /v1/ai/generate/speech endpoint can select the same underlying engine with "model": "ida-voice", but only as a fixed built-in voice — it does not accept a reference-audio upload, a saved voice profile, or an emotion vector. There is currently no public route or working MCP tool to clone a voice, compare two samples for speaker identity, or run deepfake/forensic analysis on audio. If a route for voice-clone ships, it will be documented here with real, verified paths.

What's actually available today

For speech and audio production today, use one of the following:

Capability	Endpoint	Notes
Voice Agents	POST /v1/voice/agents (see Voice Agents)	Realtime conversational agents with prebuilt voices, powered by xAI Grok Voice.
Text-to-Speech (built-in voices)	POST /v1/ai/generate/speech	model: "google", "ida-voice", or "grok". No cloning, no emotion vector.
Gemini Generative TTS	POST /v1/ai/tts/gemini/synthesize	Expressive voices with natural-language style prompting. See Music & Audio.
Azure Speech Neural TTS	POST /v1/ai/tts/azure/synthesize	Neural voices across many languages, with SSML emotion/role controls. See Music & Audio.
Amazon Polly TTS	POST /v1/ai/tts/polly/synthesize	See Music & Audio.

None of these accept a reference audio upload to clone a new voice, compare two samples for speaker identity, or detect synthetic/deepfake audio — those capabilities are not exposed on the public API today.

Example: built-in-voice TTS

Python

from fotohub import FotoHub

client = FotoHub(api_key="fh_live_...")

audio = client.request(
    "POST", "/v1/ai/generate/speech",
    json={
        "text": "Hello, this is a test of the IDA Voice engine.",
        "model": "ida-voice",
        "language": "en",
    },
)

cURL

curl -X POST "https://apis.fotohub.app/v1/ai/generate/speech" \
  -H "Authorization: Bearer YOUR_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "text": "Hello, this is a test of the IDA Voice engine.",
    "model": "ida-voice",
    "language": "en"
  }'

See Music & Audio for the full text-to-speech reference, including pricing, supported languages per provider, and voice lists.