Slides: From Transcription to Live Music: Gemini's Audio Stack — Thor Schaeff, Google DeepMind
Source Video
From Transcription to Live Music: Gemini's Audio Stack — Thor Schaeff, Google DeepMind
Relationship To World's Fair 2026
These slides are extracted from a public AI Engineer YouTube video connected to World's Fair 2026. Speaker-matched clips are supporting context unless later confirmed as exact session recordings; official livestream recordings are day-level/event-level source material.
Related Scheduled Sessions
- No individual scheduled session mapping has been assigned yet; treat this as an event livestream deck.
Extracted Slides

OCR text:
€3 Braintrust €} WorkOS OpenAl

OCR text:
a) a - , ao¥, a ma
: FENG
a ark
WHAT'S NEW IN Al AUDIO? yee <
™ Z 7
at ole at.
Google DeepMind
Dd
ra oy , |
& ;
ne 7 7 _ e
I ae
‘ ae

OCR text:
ae
a 4
Pane a .
; ww lees]
rs Fs i ka
a r a fey a
ce,
. fet
py
Thor Schaeff
Developer Relations Engineer,
ce tarosannelole lang
oy Engineering the future of Al

OCR text:
Shipping at relentless pace
FrontierModels
ni 2.5 Pro
MS'ZN
ini2.5Pro
Gemini 2.5 Flath
AIE
ni2.0
L.SFlaa
n2.0
2.0Pre
2O24May-Jun-Jul-Aug-Sep-Oct-Nov-Dec2O25Jan-Feb-Mar-Apr-May-Jun-Jul-Aug-Sep-Oct-Nov-Dec
Gemma2
Cemma3
Gemma4
TlL. 279
E2L (43: 2HE.AAL 211
Gemma3n
OpenModels
AlEngineer
EUROPE

OCR text:
genMediaModels
Veo3.1 Lihe
en4 Fast
LyrlaReafTime
Lyrla3 Clp6Pro
AIE
YD2
Veo2
/eo3.1
feo3.1Fa
2024May-Jun-Ju-Aug-Sep-Oct-Nov-Dec2O25Jan-Feb-Mar-Apr-May-Jun-Jul-Aug-Sep-Oct-Nov-Dec2O26
Jdy-e-qa-uer
Chirp3
Flash/Pro TTS
Gemii2.5
Gemini 2.S Flash Native Audic
Gemini 2.5 Flash Native Audlo
Audiomodels
Engineering the future of Al
AIEng

OCR text:
ranean’ : Our goal is to build models that
i i deeply comprehend. richly
mn . : transcribe, and robustly reason
ran through audio -- seamlessly
: handling any mix of languages,
; dialects, accents. and modalities.
Anywhere. Always.
Es
| Al Engineer |
seco) a
’

OCR text:
Pe
Turn your audio into accurate text © ee
Paka a
* oy + AR tn Oy Be
* ba Ce See ak alba rhs aa
* a : ta tf ace
Oe
* ad
a
Start Recording
Baa COTS Las]
| Al Engineer |
Sue) a 3
i

OCR text:
es on fae
‘ ra é fo 8
EchoScript ~ Remix Ca Device i?) Sb
EchoScript Al Power ty Gena A ash Prawn &
Transcription Results Start Over
Pei
bf bd
Summary
* id
aa Thor from Google DeepMind introduces the topic of Al Audio, specifically focus:ng on
developments within DeepMind and the Gem:ns API. He introduces himself and
demonstrates multi-lingual capabuities Dy greeting the audience in English, German,
French, Japanese, and Chinese
.
Detailed Transcript
& Thor Ayre 8 oe Mies © Nets
Allright, what's nev in Al audio? I'm sorry, it's a bttle bit misicading because the tie
leaves out the ‘at Google DeepMind’. So we're just kind of looking at, you know, what
Engi ing the fut a
g fe
| ate . ~~

OCR text:
Audio Understanding
constmodelId="gemini-3-flash-preview"
const prompt=
You are an expert audio transcription assistant.
Process the provided audio file and generate a detailed transcription.
AIE
Requirements:
1.Identify distinct speakers (e.g.,Speaker 1,Speaker 2,or names if context allows).
2.Provide accurate timestamps for each segment (Format:MM:SS).
3.Detect the primary language of each segment.
4.If the segment isin a language different than English,also provide the English
translation.
5.Identify the primary emotion of the speaker in this segment.You MuST choose exactly
one of the following:Happy,Sad,Angry,Neutral.
Engineering the future ofAl
AIEn

OCR text:
Prompting structure
A robust prompt ideally includes the following elements that come together to
ae craft a great performance:
. mn e Audio Profile - persona / character identity (age, background
ad Ace)
* . .
e Scene - Sets the stage. Describes both the physical
environment and the “vibe”.
e Director's Notes - Performance guidance: style, breathing,
pacing, articulation and accent.
e Sample context - contextual starting point
“) e Transcript
Ce Ba
error:
ad
fee eet na
ee
Cg
¥ a

OCR text:
Prompting structure
A robust prompt ideally includes the following elements that come together to
ak ¥ craft a great performance:
. * e Audio Profile - persona / character identity (age, background
etc)
* ve bd
e Scene - Sets the stage. Describes both the physical
environment and the “vibe”.
e Director's Notes - Performance guidance: style, breathing,
pacing, articulation and accent.
e Sample context - contextual starting point
! e Transcript
Ce a
paves
per Ena
et
CR
Engineering the future of Al
\

OCR text:
rr ~ Gemini
mn .
3: vest aucia Text, audio, video +
App Bee eee Live API
Engineering the future of Al
.

OCR text:
Yi gt}
Bi iree are CRE Taal oem OC Tema TSHeI IC)
Poi
* a Model Model ID Tai arelg ecTieeh itera) Output
sae
we * Tarn) Cast Ems et rd Spee eo et Oe eas
iol ee Cs
Lyna 3 ORT tena lai Oe ree cera Pmeet ere CSHB Eee na aL aed
isfy eed BU CrmnS ee om Mtoe: (Peres On eas Use Cee ST FeO MEE
Fae
pares
Pad
Paerior ——
| Al Engineer |
" SUL cela

OCR text:
| Al Engineer |
EUROPE
AI.ENGINEER
Slide-Derived Subjects To Review
Subject extraction uses video title, related session titles/descriptions, transcript context, and OCR text when available. OCR is best-effort and should be reviewed against the embedded slide images.