Unverified50% confidenceFactExact time
Gemma E2B 变体具备原生音频理解能力，可直接处理音频频谱特征而无需独立的语音识别模块
1
Sources
50%
Confidence
Long-term
Relevance
7/11/2026
First Seen
Sources
WisprGemma：纯浏览器本地语音输入，数据不上云
redditr/GoogleGeminiAI7/10/2026
Related Claims
UnverifiedGemma 4的1B/2B模型自带语音识别功能79% similarUnverifiedGemma 4 adds audio conversation capabilities, supports audio formats including wav, mp3, m4a, flac, and webm, and introduces a "preserved chain-of-thought" feature.67% similarUnverified自回归神经语音模型的代表包括微软的VALL-E和VALL-E 265% similarUnverifiedThe technical basis of voiceprint recognition is acoustic parameters such as vocal tract morphology and fundamental frequency features, which are considered individually unique, but voice cloning can bypass automatic voiceprint verification systems based on MFCC or deep embedding vectors.64% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/488222API
curl https://kongchang.com/api/v1/knowledge/claims/488222MCP
get_claim(id=488222)