Because all the delays are different lengths you would have to create 256 delays. You could do that with a replicator. In which case, in the Delay parameter of the delays you would put something like:
(?VoiceNumber - 1) * 10 ms
with a DelayScale of 1.
You could use a PulseTrain to select out each partial by multiplying the PulseTrain by the amplitude channel of the spectrum (the left channel) using a Product Sound. If you have a PulseTrain with a duration of 256 samp and a duty cycle of a single sample width that would select the first partial. You would then feed that into the Delay. (The PulseTrain has to be 256 samp in duration to synchronise with your spectrum frame).
To select each subsequent partial you could delay the PulseTrain before multiplying with the amplitude channel by passing it through another Delay Sound (let's call that one the "PulseTrain Delay"). The size of the delay would be:
(?VoiceNumber - 1) samp
But 256 delays may be overkill for what you're trying to do. Instead you might want to split the spectrum into a number of bands, like 16 bands. To do that you want to increase the DutyCycle of the PulseTrain so it selects more than a single sample. The DutyCycle would be something like:
16 / 256
and the PulseTrain Delay would be:
(?VoiceNumber - 1) * 16 samp
This would give evenly spaced bands. Alternatively you could use a SyntheticSpectrumFromArray Sound instead of a PulseTrain to select the different bands and then you can more easily adjust the width of each band.
Gustav made a great video describing how you do that.