Netflix Big Bug Sound Designer Théo Serror has a growing list of credits under his belt. In the thread below, Théo wrote about the process that went into the Film:

Now that Jeunet's #BigBug has been out for a bit, I figured it might fun to do a little @EssaHansen-style #sounddesign thread about it. 🧵 pic.twitter.com/hW3508tp2n
— Théo Serror (@t_serror) March 6, 2022
One of my jobs on this film was Character Voice Design.
— Théo Serror (@t_serror) March 6, 2022
There are 6 robots, 3 of which played by actual human beings (androids).
The three "mechas" were voiced _on set_ in a sound booth. It allowed for "organic" (?) performance and interactions between characters. pic.twitter.com/qGEh5DO40G
Since Jeunet's films have such a strong visual identity, sound inevitably follows that path. Meaning a lot of thoughts went into understanding "why does each robots looks like it does ?" "how does it work ?" "how is it built ?". pic.twitter.com/nOTUK100MD
— Théo Serror (@t_serror) March 6, 2022
This may seem very abstract, but it helps as sort of a Ariadne's thread. Here : each robot seem to belong to a different "Gen" (using console terminology) :
— Théo Serror (@t_serror) March 6, 2022
-Decker (vacuum cleaner) being the most basic (Amstrad CPC era)
-Greg (male android) being the most elaborate (Real Human). pic.twitter.com/Eavg01qaNQ
We tried to mimic that idea in there voice design : assigning one "technique" to each bot, forming sort of a "Voice Processing Family Tree" from good ol' Vocoder to Antares Autotune, etc. pic.twitter.com/pfQd5vvGVZ
— Théo Serror (@t_serror) March 6, 2022
E.g. :
— Théo Serror (@t_serror) March 6, 2022
Decker, being the "most vintage" one (looking like a SMEG Fridge), was 100% vocoded.
To add an extra layer of "robotness" : I revoiced all his lines feeding a synth to a talkbox, mimicking the actor's performance.
This "talkbox" track was only open on selected lines. pic.twitter.com/YzCJHVvgej
This allowed us to highlight her inner conflict by tuning the effect on a line-by-line basis. Jeunet was also really adamant to stay true to the actress's performance – which was a challenge at times.
— Théo Serror (@t_serror) March 6, 2022
To be specific, I mostly used Autotune EFX, and set the note on which her voice would tune *on a syllable-to-syllable basis*. The further the pitch, the more robotic.
— Théo Serror (@t_serror) March 6, 2022
This also created a "song" to each of her line.
(Automating that lil keyboard was painful😂). pic.twitter.com/t8nFOFCZYW
Final example : Eisntein. With his clockwork-DIY style building, it was tough to find a strong idea. Plus he has a TON of lines (often long and complicated ones too), and Dussolier's performance is full of nuances that easily get lost through processing. pic.twitter.com/o7w8JrwCdy
— Théo Serror (@t_serror) March 6, 2022
This led to to settle for sort of a DIY technique as well: using a transducer, we reamp'd his lines through an actual piece of furniture.
— Théo Serror (@t_serror) March 6, 2022
What transducers do that IRs don't is all the physical accident that happen *in reaction* to the signal (a hinge might start squeaking)…
… or the joint might resonate and rattle on some syllables, …
— Théo Serror (@t_serror) March 6, 2022
Second : it allowed us to boom this reamped lines, mimicking screenplay in a very organic way at times and play with acoustics.
A few key things I learnt from this :
— Théo Serror (@t_serror) March 6, 2022
1/ 80% of the "robotness" comes from talent's performance. This seems obvious, but it's a great way to guide the creative process. Tom & Decker's performers already had a vocoder-ish/vintage toy quality to their perf (monotonous, stuttery).
2/ Brain only need a split-second to accept an effect. This helps especially on longer, involved lines where the effect may get in the way of story delivery.
— Théo Serror (@t_serror) March 6, 2022
There, having full blown effect for a couple word at the beginning and end of the line, while remaining subtle…
…for the thick of it help with intelligibility, while allowing for strongly characterized effect overall.
— Théo Serror (@t_serror) March 6, 2022
For More From Théo: