ComingUp
Gallery
About
I have been building Speech Studio, an open-source local voice-cloning app, and just published a benchmark for the models behind the current stack. Benchmark: https://www.soniqo.audio/blog/voice-cloning-benchmarks Models tested: OmniVoice int8 Chatterbox Multilingual fp16 VoxCPM2 bf16 Fish Audio S2 Pro fp16 Languages: English, German, Modern Standard Arabic, Spanish, Mandarin Chinese. The page includes reference audio, generated audio, speaker similarity, WER/CER, generated audio length, and RTF...
Comments (9)
five languages is incredible
cross-language WER numbers would make this more useful
which of the five languages scored the worst?
five languages feels like a small sample for any real benchmarking
Bet english dominated the scores, what about the other four languages?
how much does quality drop for non-english languages
Really useful seeing these models compared across languages
how do these stack against elevenlabs pricing?
Which language had the most natural sounding clones?
Related Products
CAVN AI | AI Music Generator
CAVN AI is an innovative platform that allows users to generate music
Typist
Transcribe audio to text in seconds.
rapym
TKCORE AI
All-in-one AI platform featuring TkCore-V5.5-Pro & top LLMs
ai image prompts
Copy-paste AI image prompts that actually work.
ChatFlow
AI chatbot for your website — live in minutes