Few-Shot Voice Cloning (2D): Speaker Embedding Plane & Vocoder Conditioning

Modern voice-cloning tools don't need hours of studio audio — a handful of short clips is enough to compute a speaker embedding that a vocoder can condition on. This 2D simulator plots a pannable, zoomable embedding plane with four distinct speaker clusters: pick a target, choose how many reference "shots" to record and how noisy the microphone is, and watch the clone embedding — the mean of those noisy reference points — converge toward the true speaker centroid as the law of large numbers averages the noise away. A conditioning-strength slider then interpolates a simplified vocoder's output formant bar chart between a generic average voice and the fully cloned identity, while live readouts track cosine similarity, embedding error and the resulting estimated pitch.