Unverified50% confidenceTradeoffExact time
ElevenLabs' similarity_boost parameter (range 0–1) controls how closely the generated result aligns with the original voice clone sample; values that are too high can sometimes produce artifacts or a degradation in audio quality.
1
Sources
50%
Confidence
Medium-term (~90 days)
Relevance
10/11/2026
First Seen
Valid until: 1/9/2027
Sources
Related Entities
Related Claims
UnverifiedElevenLabs' stability parameter (range 0–1) controls the consistency of generated speech; higher values produce more stable output with less emotional variation, making it suitable for broadcast scenarios.76% similarUnverified自回归模型（如VALL-E）逐token预测语音离散表征，能建模长程依赖和丰富韵律，但存在累积误差；非自回归模型（如FastSpeech）并行生成速度快但表现力受限69% similarUnverifiedAfter completely excluding 10 sound categories from training and then testing on them, those unseen sounds showed no improvement, indicating that the model learns given labels rather than general auditory capability and does not automatically generalize to untrained labels.69% similarUnverifiedFull multimodal large models that support voice input and output score far lower on agent benchmark datasets than mainstream reasoning models.68% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/998615API
curl https://kongchang.com/api/v1/knowledge/claims/998615MCP
get_claim(id=998615)