A new platform called TTS.ai is packaging over 20 open-source text-to-speech models into a single hosted service, adding voice cloning, transcription, and audio processing tools on top.
The pitch is consolidation. Instead of self-hosting Kokoro, CosyVoice 2, StyleTTS 2, or Tortoise TTS individually, TTS.ai runs them all on dedicated GPU servers and charges by credits. Free accounts get 15 credits with access to 4 basic models. Paid plans start at $9/month for 500 credits and scale to $99/month for 10,000 credits, with API access (OpenAI-compatible) kicking in at the $29/month Pro tier.
The voice cloning feature needs just 5 seconds of audio to generate a clone, which is competitive with ElevenLabs and similar services. The platform covers 107+ voices across 32 languages, outputs in WAV format with conversion to MP3, FLAC, and others, and all underlying models use permissive licenses (MIT, Apache 2.0) that allow commercial use without royalties.
The operator here is Muddy Holdings LLC, not a name with existing traction in the AI voice space. That matters because reliability and uptime are the whole value proposition of a hosted wrapper around open-source models. Anyone technical enough to know what these models are could self-host them for free. You're paying for the convenience of not managing GPU infrastructure, and that only works if the service stays online and keeps its models current.
For teams already using ElevenLabs or LOVO, there's no obvious reason to switch. The model variety is interesting for experimentation, but production voice work typically needs consistency, not a buffet. Where TTS.ai could find an audience is among developers who want a single API endpoint to test multiple TTS engines before committing to one, or small creators who want voice cloning without ElevenLabs pricing.