Quantcast
Converting a Real Person Into Siri - Gearspace.com
The No.1 Website for Pro Audio
Converting a Real Person Into Siri
Old 25th September 2012
  #1
Lives for gear
 
🎧 10 years
Converting a Real Person Into Siri

Hey, all. Posting a commercial that involves an actor's iPhone giving her advice and pitching a retailer. The 'Siri' voice is the retailer's normal spokesperson, so we need to maintain her identifiability while also giving the impression her voice is computer-generated. I've found several ways to effectively turn her into a monotone robot, but haven't found a combo that will do the digital thing and still keep her inflections intact. Any suggestions?
Old 25th September 2012
  #2
Gear Addict
 
monsieur x's Avatar
 
2 Reviews written
🎧 10 years
The Vocal Transformer (or whatever it's called) in Logic could maybe work.

Or, this is crazy:

use one of the many voices that come with OS X, (i.e. what happens when in terminal you type in say hello -v vicki)

You can find the voices under the Speech (Text to Speech) preferences.

Take that audio, and vocalign or cut/stretch/slide/move to layer it in under her voice. (Could sound horrible.)

Anyway, hope there is something helpful here,

Good luck,
Old 25th September 2012
  #3
Here for the gear
 
🎧 15 years
I would check out melodyne. It's really easy to make something very close to the voice in the computer game Portal.
Old 25th September 2012
  #4
Lives for gear
 
Rob King's Avatar
 
🎧 15 years
Before you do any processing, have her read it like 10 times, then cut up words from each take and put them all back together randomly to make the paragraph or phrase. Ghetto Phonetics.. Then process it. It kinda works.
Old 25th September 2012
  #5
Lives for gear
 
🎧 10 years
It's a bit long winded but here you go:

Old 25th September 2012
  #6
Lives for gear
 
🎧 10 years
Thanks all, good stuff. We did talk about having her read the copy one word at a time, but decided against it. I do have Melodyne, so I think I'll go that route. I don't use Logic for post work, but I do have it as a fallback.
Old 25th September 2012 | Show parent
  #7
Lives for gear
 
Etch-A-Sketch's Avatar
 
🎧 10 years
Quote:
Originally Posted by seanmccoy ➑️
Thanks all, good stuff. We did talk about having her read the copy one word at a time, but decided against it. I do have Melodyne, so I think I'll go that route. I don't use Logic for post work, but I do have it as a fallback.
Really it's all about delay, feedback loops, ring modulators and maybe throwing in a vocoder as well (although it's not needed). People were making voices sound computerized long before Melodyne. Transformers is a perfect example of that.

Look into how the voices for transformers (original cartoon and new films) were done and that will give you your answer.
Old 25th September 2012
  #8
Deleted e479b20
Guest
speakerphone has celp and gsm codecs. although siri sounds far from robotic IMO. more a question of using a natural voice and chopping up the words from lines with other meanig to the one you need. more than actually making it a robot voice from transformers.
computer voices are not what they used to be ;-)
Old 26th September 2012
  #9
Gear Nut
 
DrummerMan's Avatar
 
🎧 10 years
Just an idea from a non-post person here, but have you tried using the robotic monotone sound you got in parallel with the original unaffected voice? Won't necessarily sound like Siri but would probably retain the recognizability of the voice while making it obvious that we're supposed to think its computer generated.
Old 26th September 2012
  #10
Lives for gear
 
🎧 10 years
To me this is more about performance. Take the voice and get them to record the words you need seperately, one at a time numerous takes of each one. then splice together. the thing about Siri is that he/she/it sounds lifeless, just a series of words one after another. there's no pre-recorded sentences, therefore no performance.

The actual tone of Siri, to me is more human than robot sounding. It's just that it sounds a little telephone filtered because of the telephone speaker it speaks from. Maybe just a touch of speakerphone to help....
Old 27th September 2012
  #11
Gear Nut
 
Subsonic808's Avatar
 
1 Review written
🎧 10 years
It definitely is all about the performance. There is a fairly recent episode of 'The Big Bang Theory' where Raj gets an iPhone and falls in love with Siri. At the end of the episode, he meets Siri in person in a dream sequence. Just watch this and you will know what you need to do. The actress who plays Siri is doing an awesome job.
Old 27th September 2012
  #12
Lives for gear
 
Etch-A-Sketch's Avatar
 
🎧 10 years
Apple Tried to Silence the Voice Behind Siri | Fox News

Quote:
Briggs made recordings of his voice years ago to a company called Scansoft, now a division of speech-to-tech company Nuance, which named him "Daniel." Nuance works with Apple on Siri, the paper noted. Briggs was unphased by the discovery that the massive technology company had co-opted his voice for its latest smartphones -- a fact he learned of only when he saw a demonstration of Siri on TV.

β€œI got paid a decent sum by Scansoft,” he said. β€œI love Apple’s products and I think Siri is a game-changer.”
Small world. Scansoft/Nuance was original a company called Dragon Systems. I was a sound engineer at dragon systems for three years before moving to LA. I worked on a lot of those types of recordings they are talking about in the article (the one that Briggs did).

Those recordings are used to create text to speech algorithms. We would basically recreate their voice using synthesis after analyzing their recordings.

Unfortunately it's a very involved process to make a truly synthesized voice like siri from one person's voice. Short of that... the old traditional routes that have been mentioned (ring modulators, filters, vocoders, and delays) are your best route as well as having the actor work on their performance to mimic lilt in the speech because of the way the text to speech engine organizes and links the phonemes.
Old 27th September 2012
  #13
Gear Nut
 
🎧 10 years
I don't have an iPhone with Siri, but is it possible to simply read the script to Siri as a "memo" (in sections?) and then just have her repeat it back? Then you could just take a line from the phone and record that?
Old 27th September 2012 | Show parent
  #14
Lives for gear
 
🎧 10 years
Quote:
Originally Posted by Allaboutrhythm ➑️
I don't have an iPhone with Siri, but is it possible to simply read the script to Siri as a "memo" (in sections?) and then just have her repeat it back? Then you could just take a line from the phone and record that?
I don't have an iPhone either so I don't know, but it's an interesting idea. In this case, though, we didn't want the Siri voice but the voice of the spokeswoman. She did an excellent job of mimicking Siri's slightly off inflection while still sounding like herself and selling the product. I ended up doing a bit of pitch and formant alteration in Melodyne and ran it through a Speakerphone small radio IR and it worked well.
πŸ“ Reply

Similar Threads

Thread / Thread Starter Replies / Views Last Post
replies: 23 views: 13473
Avatar for matt-o-
matt-o- 16th September 2011
replies: 3 views: 2657
Avatar for mekohler
mekohler 22nd May 2013
Topic:
Post Reply

Welcome to the Gearspace Pro Audio Community!

Registration benefits include:
  • The ability to reply to and create new discussions
  • Access to members-only giveaways & competitions
  • Interact with VIP industry experts in our guest Q&As
  • Access to members-only sub forum discussions
  • Access to members-only Chat Room
  • Get INSTANT ACCESS to the world's best private pro audio Classifieds for only USD $20/year
  • Promote your eBay auctions and Reverb.com listings for free
  • Remove this message!
You need an account to post a reply. Create a username and password below and an account will be created and your post entered.


 
 
Slide to join now Processing…

Forum Jump
Forum Jump