GPT-4o voice mode is really impressive.
We gave it a face with
@synthesiaIO
EXPRESS-1, our latest avatar model.
When empathy is important – healthcare, coaching, education - a friendly face really makes a difference.
It's why we do Zoom over phone calls.
What do you think?
Excited to announce our $90m Series C at $1bn valuation led by Accel w/ participation from Nvidia.
In 2017, when we founded Synthesia, generative AI sounded like sci-fi to most.
It maybe also did for us; but we kept building. And here we are in numbers, 6 years later...
AI Avatars aren't great actors...
They don’t actually understand what they’re saying or how they should say it.
Until now – with EXPRESS-1, we’ve trained our avatars to perform like an actor: tone of voice, facial expressions and body language👇
Truly a moment of Synthesia history: I’m in a suit 🫡
Thank you for the 10 Downing Street invite
@RishiSunak
and excited to play our part in making the UK an AI powerhouse!
Synthesia just featured on 60 Minutes 🤯
Bill Whittaker now has his own AI avatar. Can you tell? 😉
So much fun to work with the awesome 60M team,
@NinaDSchick
, Descript & more to tell the story of synthetic media.
Link to full piece below 👇
Big milestone
@synthesiaIO
: Raised a $50m Series B led by Kleiner Perkins w/ follow-on from Google Ventures.
Just getting started building the world’s best AI video platform!
Now, we'll accelerate our progress towards video production w/o cameras, studios or actors.
👇
Our first real Synthesia ad 🤯
Time to make your own AI avatar and get rid of those steaming piles of text... 📝
Finally able to explain to my grandma what this AI Video stuff is all about. Huge! 👏
We just launched
@SynthesiaIO
2.0, the world’s first video communications platform built for businesses ✌️
- New features like AI screen recorder and interactivity
- Improved creation & translation workflows
- Next-gen AI avatars (soon with hands!)
More details in 🧵
We just launched the world’s most realistic Personal Avatars at
@synthesiaIO
👇
Looks like you, sounds like you and speaks 29 languages fluently.
Record in 5 min w/ webcam or smartphone – you can get creative!
Check them out and try for yourself:
Synthesia STUDIO 2.0 is finally here, with a complete redesign for improved look, functionality, and performance 🎉
Over the last year, we've been inventing new UX paradigms to capture the potential of AI-first video generation. This is just the beginning.
Still no timeline 😉
In 5 years most media content will be synthetic, and by far the majority of it will be positive. But there will be challenges.
Here's our vision and how we're building synthetic media tech to maximise human creativity and minimise harmful use of AI.
Three days into Davos, it’s striking how much the EU AI Act is referred to as a ‘competitive advantage’ by everyone else than those building AI companies.
If EU wants strong regulation, it will need to counterbalance with actual competitive advantages: access to public datasets,
- 90% of the realism is due to this being a TC impersonator in the first place (hair, voice, clothing, acting).
- Faceswap is the cherry on top that makes it feel *almost* uncanny. It corrects for the only thing the impersonator can’t fake convincingly: the face appearance.
1/ Deepfakes, deepfakes, deepfakes...
99% of media coverage is still negative.
Inversely correlated with actual negative use, which is likely less than >0.001% [1]
Let's celebrate the real business value synthetic media is generating today.
8 high-growth applications👇
Next-gen AI avatars - now with hands 🙌
Full-body, fully controllable AI avatars that can interact with their environment and gesticulate with their hands — coming to Synthesia later this year.
- Software used here is probably open source. Likely tweaked by the creator.
- Taking on/off hat and sunglasses is very cleverly used as transition point from real video to deepfake. Video is only synthesised in the beginning/end, not in the middle when he’s swinging the club.
If you could generate infinite video with AI, instead of cameras, what'd you do?
We made Anna, our AI avatar, count to 1 million live on TikTok 🙇♀️
It'll take her 45 days & she's doing live QA on stream. Ask q's and I'll pass em on 😉
9m views so far 📈
🎉 50.000 customers
🎥 15m videos generated
🙌🏻 More than 200 Synthesian's making it happen
📈 Sustainable + rapidly growing business
More thoughts & coverage below 👇
The OpenAI board has 3 independents + Greg Brockman, Ilya Sutskever and Sam Altman.
So presumably either Ilya or Greg must have voted against Sam. This isn’t VCs seizing control. Wild.
Most tech teams today are bogged down by never-ending processes & meetings.
At
@synthesiaIO
we flipped the model for our product team:
- 1 recurring meeting / week
- No non-tech PMs
- No Jira
- Super small teams
- Progress > perfection
👉Want to join our eng team? DMs open!
Really proud of the Synthesia team for pulling off our first SIGGRAPH paper!
HumanRF is our first step towards next gen avatar tech that will enable much more rich and complex AI video. More magic coming soon 🔮
For the first time ever we're also opening up some of our
🎉 Excited to share HumanRF
#SIGGRAPH2023
🎉
It compactly reconstructs full humans in action and enables novel-view synthesis. Check out the new high-quality dataset ActorsHQ too!
Code and data follow soon.
Project page:
Paper:
AI avatars with emotions, expressions and acting skills are now live. Built on our EXPRESS-1 model.
Free to try, link below – use emotional scripts for best effect 😉
Happy to see reason prevailing despite the AI mania.
It's key to target harmful outcomes, not specific algorithms or technologies.
In the short term AI will heighten existing issues like fraud, misinformation, and privacy. Regulators first step should be to adapt existing laws
- This tech will be a game changer for celebrities in terms of scaling their likeness/IP.
- Brands can work with impersonators on set and finalize content by having the celebrity voice-over the final content (until voice cloning is good enough).
- While this is still *very hard* to pull off (and still would require impersonator, studio etc) someone with enough resources can certainly use this maliciously.
- My bet is that 100% of publicly available deepfake detectors will not be able to detect this.
Text-to-speech will likely be a solved field in 6-9 months. We’re way past the uncanny valley and the tech is commoditising rapidly.
Main vectors of improvement from here are stability, phonetic control, style prompting and expansion to all world languages.
So many cool new
Use your voice to engage in a back-and-forth conversation with ChatGPT. Speak with it on the go, request a bedtime story, or settle a dinner table debate.
Sound on 🔊
By watching many, many hours of people speaking, our latest model has learnt to replicate the tiny micro expressions and body language that humans make.
If their script is sad they’ll look and sound sad. If they’re excited they look upbeat and energetic.
1/ A lot of talk about how GPT will 'replace' search.
Feels less known (?) but GPT is already massively impacting search.
GPT co’s like Jasper (~$80m ARR) make all their $ from mass-producing SEO content.
Will GPT beat Google Search at it's own game and then replace it?
👇
PMF & longevity test for GenAI companies:
Do your customers overwhelmingly present what they make as *AI-generated* text/image/video/etc?
If the answer is overwhelmingly yes, you are likely monetizing novelty, not utility.
1/7: AI apps are experiencing high growth and high churn, but these are closely related. The abundance of new apps creates excitement, but eventually, the party will end.
- This is clearly a very experienced deepfake and visual effects artist(s).
- Don’t underestimate how much time and resource it took to make this. Likely weeks.
- Maybe the same group that did this:
This is only the start and it’s a game changer.
Expressive Avatars launching this Thursday, April 25th on this event:
And we're hosting a pre-launch live chat on X with
@MattNiessner
,
@jnstrck
,
@alexvoica
now!
Join here:
This is a completely generated AI Avatar of Lionel Messi.
Messi mania is in full swing and we reminisced about the legendary collab we did at
@synthesiaIO
2 years ago.
Fun fact: More than 10m people generated messages for their friends 🎉
Thanks
@Jason
for having me on the pod! 🙌🏻
Loved this segment on how we implement AI safety and how we think content moderation increasingly will happen at the point of creation, not just distribution.
Full pod:
.
@synthesiaIO
's software enables anyone to create AI-powered videos with synthetic avatars
according to
@Jason
, Synthesia's biggest issue: bad actors using it to create harmful deepfakes
CEO
@vriparbelli
explains his responsibility and tactics for stopping this problem:
We're taking Synthesia to the next level! 📈
Looking forward to collaborate with a fantastic group of investors to execute on our vision to make content creation easier and faster using AI.
We won a Cannes Lion 🦁
Proud of the whole
@synthesiaIO
team and a big win for synthetic media.
We got a lot more celeb stuff coming this year. As always pushing the boundaries of what's possible.
Prepare to get mind blown 😉
Two avatars in GPT3-driven convo about humans & AI live on TikTok w/ millions of viewers. This is their touching goodbye 🥹
What a beautiful moment in time & a precursor to the cambrian explosion of creativity we're about to witness.
❤️ my team and all the crazy shit we do
Delivering this Beckham campaign using Synthesia technology feels like the culmination of two years of hard work 👏
Really proud of the Synthesia team and everyone else involved in bringing this amazing campaign to life!
More than 1m views of David Beckham using Synthesia to create awareness about Malaria. We’re really proud to have been involved in this campaign. Let’s end Malaria once and for all!🙏 -
My 5 Predictions for Synthetic Media in 2020:
🎬 Synthetic Media in media production.
🗣️ Synthetic voice will cross the uncanny valley.
✅ Ethical standards will emerge.
📱 Consumer apps on the rise.
🇺🇸 Deepfakes won’t impact 2020 election.
I’m excited to attend the
@WEF
’s annual meeting in Davos this year, representing Synthesia & the AI industry.
2024 will be an important year for AI – real-world applications will impact almost every business in the world directly or indirectly. So we need to get it right.
I (genuinely) wanted to do record a video update going through all the new amazing things we've shipped.
Served as a good reminder why we founded
@synthesiaIO
... even simple video production sucks 😉
This week, I was invited to speak
@CommonsBTC
about how the right policy can build on the UK’s strengths and competitive advantages in AI – GPUs, data and mandated public sector adoption:
For
@fortunemagazine
:
Great to see the UK paving the way on sensible AI regulation that protects victims of online abuse.
Sharing non-consensual intimate deepfakes will now be criminalized and punishable by up to two years in prison as per the new Online Safety Bill.
We introduced a 4 week paid sabbatical + $10K after four years tenure
@synthesiaIO
.
Suggestion from an employee and when I heard it, I loved it.
Pursuing our ‘one day…’ dreams shouldn’t be incompatible with work (Kilimanjaro👇).
What are other awesome, non-obvious perks?
Media: DEEPFAKES CAN BE CREATED IN SECONDS BY ANYONE WITH ACCESS TO A COMPUTER RUNNING WINDOWS 95 OR ABOVE
Also the media: Pays Framestore, one of the world's top 3 Hollywood visual effect studios, big $$$ to manually create a ‘deepfake’ on a film-quality green screen stage.
🤔
In the early days of the internet we had a username. Then we had a profile picture. Social profiles. Your own 2D bitmoji. The latest iteration is the 3D Memoji from Apple. Next up - your own high-fidelity AI avatar.
@patrick_oshag
Same, but
@synthesiaIO
not Unreal...😉
We need AI native rendering engines that remove all abstraction layers (like DALL-E) so that users can interact with them as a creative directors, not as craftsmen.
The next Golden Age of media will be even wilder than we think.
AI generated Messi videos -- very proud of this collab with
@PepsiCo
,
@LAYS
and one of the greatest footballers ⚽️
Last few months have been crazy for AI assisted content creation.
This is essentially scaling the
@BookCameo
experience.
We've still only scratched the surface📈
⚽️ Lionel Messi now speaks the language of his superfans!
💥 Today we can finally share our latest collaboration with
@PepsiCo
,
@UNIT9
&
@180Amsterdam
.
🎥 Our technology enabled Leonel Messi to speak 8 languages.
[New on the pod] This week’s episode is with
@vriparbelli
, CEO of newly-minted AI unicorn
@synthesiaIO
🦄
* Starting a Generative AI startup years before Generative AI was cool
* Tactical lessons learned building a successful AI startup
* more!
All links in the next tweet👇
Thanks to
@MayorofLondon
and
@MMC_Ventures
for a great event yesterday. Proud to be a Londoner and incredible to see the public and private sector collaborating on supporting the startup ecosystem!
Start-ups and small businesses are the lifeblood of London’s economy - so it’s vital they have the best chance to achieve their ambitions.
Proud to launch our new £100m Greater London Investment Fund to help small business owners in our capital.
WIRED on how 30 Ernst & Young partners now use
@synthesiaIO
with their own AI avatar.
Love seeing how much value AI video are generating for EY and their clients. More to come!
Thanks to WIRED and
@tsimonite
for covering!
Synthetic media is the future of content creation.
BBC interviewed me for my take on where we're headed, beyond the media headlines 👇
cc:
@BernieDebusmannBBC
1/3 - Our latest work with the BBC: Live weather report by a photo-realistic AI-avatar. Pick a city in the drop-down to try it out.
🌦️ Weather data is pulled live from the BBC API and videos are generated on-the-fly using the Synthesia API.
Some thoughts on why I think
@sanalabs
is an exciting company in the 'What to Watch in AI' series.
Fun to write about something else than Synthesia for once!
Thanks for letting me contribute
@mariogabriele
🎉
AI-generated content is a paradigm shift in media production. And it's already here.
Join us in designing the UX for the next generation of creators with us at
@synthesiaIO
.
Looking for a Lead Product Designer. Can be based anywhere 🌍
Our portfolio company,
@synthesiaIO
is seeking a Lead Product Designer based in London, great role as strong dev team and 1000s of users provide feedback on a daily basis -->
Congrats
@matistanis
- huge win for 11Labs & for London!
Not a Brexiteer by any means, but UK sovereignty is a huge advantage in building a strong AI ecosystem.
Only place to build in Europe without being sucked in the EU’s anti-innovation vortex.
First episode of our new
@lightspeedvp
podcast Generative Now is live! Had a great conversation with
@vriparbelli
, Co-founder/CEO of generative video startup
@synthesiaIO
.
Topics we discussed:
- Every AI company’s growth is exploding
- AI is replacing traditional video
It’s pretty simple: you take a real video, loop it and use AI for lipsync + the mouth region with 11 Labs voice cloning.
Body language, hand/eye movements and the background is fixed to whatever was recorded which is why they still have an uncanny quality to them.
For this to
So the "AI generated girl" in the car video?
Yeah, she appears to be a 100% real person named Arial who does gig videos on Fiverr
Some of the biggest AI influencers on X have been telling everyone this was AI generated, and two different AI video companies including
Honoured to have made this years Forbes 30 Under 30 List and incredibly grateful for all the amazing colleagues, investors, clients and friends/family that have made the journey possible through the ups and downs 👏
The team behind
@SynthesiaIO
has pushed the boundaries of what’s possible in
#generativeAI
with precision and passion.
@vriparbelli
discusses
#AI
and the transformation of video production, research, and regulation with
@pbotteri
in this episode:
I'll be on the
@ClamanCountdown
with
@LizClaman
to talk about Synthesia's new AI avatars and the platform we're building to help businesses transition to a video-first world. Join us today at 3.25pm ET on Fox Business!
2/ This year we've seen the adoption of AI video skyrocket.
We've generated more than 6 million unique videos for our customers.
How does it work?
You create videos by simply typing in the text - no cameras/studios/actors required.
Happy to announce that
@synthesiaIO
is moving out of stealth. We do exciting native movie dubbing with cutting edge deep learning technology, empowering storytellers with AI :)
No more bad lip sync on TV!
#deeplearning
#vfx
#AI
Interactive videos🎮
Create rich video experiences with features such as clickable hotspots, embedded forms, quizzes, and personalized CTAs for an entirely new video experience.
Excited to announce our $13.5M Series A round at
@softr_io
led by
@mattturck
at FirstMark Capital, and 35+ amazing angels, the world’s top minds in product, growth and community. 🎉🚀
We are building the world’s largest ecosystem for no-code apps.
5/ Sneak peek: we are now building a London based studio to capture detailed 3D human data at scale.
One step closer to replacing cameras & physical sensors with software.
More coming soon!
Excited to be giving a talk around synthetic media at Facebook HQ for the next
#LondonAI
event on 28th March.
Will be in truly great company with talks from Peter Ondrúška of Lyft, Bora Edizel of Facebook AI and Petar Tsankov of ETH Zurich.
Sign up:
What is it like to be a computer? Phenomenal work by
@_ericelliott
using Synthesia and GPT3 with almost half a million views. Incredible how far we come & how far we still are from real computational intelligence.
AI Video Assistant - bulk video creation🔗 + brand kits
Soon, our users will be able to simply select a branded template, provide a link to their knowledge center, and the AI video assistant will transform all the articles into short, engaging videos.