What is OmniVoice?
OmniVoice is an open-source AI voice generator that converts text to speech, clones voices from short audio samples, and creates custom voices from text descriptions across 646 languages. The model can be self-hosted under the Apache 2.0 license or accessed via a hosted web service using one-time credit packs.
What are the features of OmniVoice?
- Multilingual text-to-speechGenerates audio from typed text in 646 languages using a single unified model, with adjustable speaking speed from 0.5× to 2.0×.
- Zero-shot voice cloningReplicates a speaker's voice from a 3–25 second PCM, WAV, MP3, FLAC, or OPUS reference clip without training, supporting cross-lingual generation.
- Text-based voice designCreates a custom voice profile from a written description of attributes like age, pitch, accent, and style, requiring no audio input.
- Expressive speech markersRenders non-verbal sounds such as laughter or sighs when inline tags are added to the input script.
- Automatic transcriptionTranscribes uploaded voice cloning reference clips using Whisper ASR.
What are the use cases of OmniVoice?
- Audiobook narrationProduces long-form spoken audio for books and stories using generated or cloned voices.
- Game character dialogueGenerates dynamic NPC voices for video games.
- Language learning toolsProvides clear pronunciation examples across supported languages for educational content.
OmniVoice Pricing Information:
- Free trial$0; provides 2 hosted credits (approximately 200 characters or 16 seconds of speech) after sign-up, with no credit card required. Includes access to all 646 languages, voice cloning, voice design, and MP3/WAV export.
- Basic$9.90 one-time purchase; provides 800 credits (approximately 80,000 characters or 1.8 hours of speech). Credits never expire.
- Pro$29.90 one-time purchase; provides 3,000 credits (approximately 300,000 characters or 6.7 hours of speech) at $0.010 per credit. Includes access to the latest voice model. Credits never expire.
- Business$49.90 one-time purchase; provides 6,000 credits (approximately 600,000 characters or 13.3 hours of speech) at $0.008 per credit. Includes the latest voice model and priority generation. Credits never expire.
- Self-hostedFree under the Apache 2.0 license, which permits commercial use. Requires user-managed compute on NVIDIA GPU (CUDA 12.8), Apple Silicon, or CPU.
