
Audio gen lead @GoogleDeepmind (Gemini 3.8 TTS) Previously led research @hume_ai (EVI3/2/1, Octave)
Joined June 2020
- Tweets33
- Following146
- Followers112
- Likes96
Amazing research by @garygarywang @kastnerkyle ! Exploiting the multimodal capabilities of Gemini.
Did you know you can generate custom voices right in @GoogleAIStudio?! Here some examples:
🇮🇪 Cork pub storyteller
🇦🇷 Buenos Aires radio host
🇯🇵 Tokyo game show host
Which one's your favorite? 👇
I asked Opus 5.5 to write, produce, and create me an educational 'cartoon' about The History of Light - using whatever resources it had available to it.
I didn't expect it to make a 36 minute video and spend $200+ dollars though 😂
The study of light is genuinely fascinating, and we have spent thousands of years trying to understand it. This is that story.
Opus 5.5 wrote and produced the entire film. I told it to use any keys in my keychain to generate the voices, create any images or video it might need, and to create whatever motion graphics it could to explain concepts in electromagnetism.
It used
> Gemini 3.8 Flash TTS for the narration, and I must say, it did an excellent job.
> Lyrica 3 for the background music, which is also essentially perfect.
> GPT Image 2.5, Nano Banana Pro and NB2 Flash for some of the scenes imagery
It's certainly not perfect but I was still really impressed with how well it did producing the film. It's very interesting if you're a physics nerd like me.
Opus is genuinely creative in a way that I've not really seen with other models. I just gave the model a few constraints and just let it make all the decisions end to end.
It even wrote itself credits at the end lol.
It's going to be so fun to create bespoke educational videos like this for my children to learn from
Thrilled to see Notebook powered by our latest Gemini 3.8 TTS! It’s been an incredible experience collaborating with the Gemini Notebook team to bring this to life.
We're pushing the boundaries of natural multispeaker dialogue. Much more to come!
Moses Oh retweeted
Interested in voice design?
Try out our latest Gemini 3.8 Flash TTS, you can create bespoke voices from scratch using natural language — customizing the role, accent, personality, and voice characteristics you want across 100+ languages and dialects.
🐉 A dramatic, fire-breathing dragon?
🎧 A charismatic narrator with a distinct regional cadence?
✨ A recognizable voice for your brand?
Congrats @AlanCowen @MosesOh4 and team for this great launch!
introducing Gemini 3.8 Flash TTS and Gemini 3.8 Flash-Lite TTS, our most expressive audio generation models yet
these models enable creators, developers, and enterprises to create richer, more expressive audio experiences
try them via the Gemini API and in AI Studio: aistudio.google.com/generate…
I’m excited to share what I’ve been working on since joining Deepmind in January!
Proud of what the team built with our latest Gemini audio gen model!
Native multi-speaker is a massive challenge. Being able to natively laugh together, backchannel, and overlap speech unlocks a huge step forward - from TTS and NotebookLM dialogs to (soon) real-time assistants.
We have a strong foundation in audio gen and looking forward to sharing upcoming models :)
Moses Oh retweeted
Introducing Octave 2: our next-generation multilingual text-to-speech model
What’s new:
- Fluent in 11+ languages
- 40% faster (<200ms latency) & 50% cheaper than Octave 1
- Multi-speaker conversation
- More reliable pronunciation
- New voice conversion & phoneme editing capabilities
For the month of October, we’re offering 50% off our Creator plan - use code OCTAVE2 at checkout!
Moses Oh retweeted
Excited to share my co-first paper at @Nature! I would like to thank my co-first authors Jason and Heydar and Prof. Dan Dombeck for his amazing mentorship. I hope that this work can open up future investigations of the underlying neural mechanisms of representation drift!
Excited to share my co-first paper in @Nature with Jason Climer and @daniel_jun_oh, under the amazing supervision of Prof. Daniel Dombeck:
nature.com/articles/s41586-0…
[1/10] I summarize our key findings in this thread.🧵
create your own clone :)
demo.hume.ai/
Moses Oh retweeted
2024: Voice Cloning
2025: What about personality cloning?
Hume’s voice AI can now not only mimic your voice but also speaking style and language.
It’s now available via our TTS and new speech-to-speech model, EVI 3, which is also launching today.
Moses Oh retweeted
.@hume_ai's @gagnecr13 & @inafried test out Hume AI's EVI 3 AI model that's optimized to understand and express human emotions at #AxiosAISummit
We released our new conversational speech-llm. Fully multimodal (inputting/outputting text&audio tokens). Ability to adopt any voice, modulate speech, and generate language.
Meet EVI 3, another step toward general voice intelligence.
EVI 3 is a speech-language model that can understand and generate any human voice, not just a handful of speakers. With this broader voice intelligence comes greater expressiveness and a deeper understanding of tune, rhythm, timbre, and speaking style.
Moses Oh retweeted
Grateful to train with awesome chiefs like @AF_Haddad. Come check us out at @NeurosurgUCSF ! #subi #neurosurgery
Applications are open for our sub-internship in neurosurgery! The sub-internship offers fourth-year medical students a hands-on experience learning about all aspects of the neurosurgery service. Apply now:
meded.ucsf.edu/current-stude…
#MedTwitter #MedEd #neurosurgery

ALT Graphic showing a photo of UCSF cerebrovascular neurosurgeon Ethan Winkler, MD, PhD, talking to a medical student during a surgery for the removal of a brain arteriovenous malformation (AVM). Banner says: Neurosurgery Sub-Internship. Text below says: First rotations start in June. Apply Before April 21.
we have NotebookLM at home 🥲
Replying to @hume_ai
Multiple interacting AI speakers:
OCTAVE can generate multiple, consistent AI personas within a real-time conversation
Excited to share what I've been working on! Over the past few months, we’ve been aligning language and speech both semantically and non-semantically in a fully trained multimodal model, while maintaining language capabilities.
It was great running into old Magenta folks at Neurips! @jesseengel @jpgard @wilzh40 @douglas_eck @GoogleMagenta
















