Evidence · first-party tested

Voice matching stayed poor in both cases, with similarity only about 15–20% on the Hindi sample and about 10–15% on the noisy sample, and the cloned voice still lost most of the source speaker’s vocal identity.

⚠ StruggledVocalAI
The observation

Voice matching stayed poor in both cases, with similarity only about 15–20% on the Hindi sample and about 10–15% on the noisy sample, and the cloned voice still lost most of the source speaker’s vocal identity.

Criterion: Voice match accuracy ·
Query this
get_evidence({
  tool: "vocalai"
})
MCP · mcp.aidemos.com/api/mcp
Free with attribution.
Real outputs, no retouching · every cell queryable via API & MCP · aidemos.com