a In vivo experimental setup for current-steering-evoked fictive singing. The nanoclip was implanted on the right-side nXIIts of an anesthetized adult male zebra finch, and an air cannula was placed in the abdominal air sac (see “Methods”). The respiratory system was pressurized by flowing warmed and humidified (35 C; 80%) isoflurane dissolved in oxygen through the cannula (0.1–0.2 L min−1 at 0.5–1.5 kPa). Multichannel stimulation (100 biphasic current pulses at 1 kHz, 200 μs phase−1) with unique spatial patterns was applied to the nXIIts with the nanoclip interface, eliciting audible vocalizations that were recorded via a microphone. b Representative examples of current-steering-elicited fictive vocalizations. For each pairing, the inset (upper right of each spectrogram) identifies the stimulation pattern that produced the vocalization shown in the spectrogram. Each spectrogram depicts the vocalization from a single trial; the orange triangle marks the onset of stimulation. c Two-dimensional t-SNE embedding of high-dimensional representations of fictively produced vocalizations (i.e., spectrograms) from one experiment testing 24 distinct current-steering stimulation patterns. Data points correspond to individual stimulation trials; colors denote current-steering patterns (see Supplementary Fig. 7 for pattern definitions). d (Left) Mean acoustic similarity between fictive vocalizations within and across multichannel stimulation patterns for n = 6 birds. (Right) Mean acoustic similarity between natural vocalizations within and across song syllable types for n = 6 birds. In each, data points show mean across vocalizations within a bird. Bars show mean across all birds in each experiment; error bars indicate sem. Two-tailed, paired t-test, within: P = 3.5 × 10−5; across: P = 0.0016. **P < 0.01; ***P < 0.001 Source data are provided as a Source Data File.