A look at how AI companies like Synthesia are creating corporate-friendly uses for deepfakes, like creating multilingual, personalized training videos
Tom Simonite / Wired : Tweets: @hollyhliu , @cidseasu , @nzaoui , @lishali88 , @far33d , @chrismessina , and @nxthompson Tweets: Holly Liu / @hollyhliu : This will change the entertainment and fashion industry @Rosebud_AI https://twitter.com/... @cidseasu : #AI professor @rao2z says the technology is impressive but wonders whether some clients may use diverse, synthetic models in place of real people from minority communities. https://ow.ly/... Noureddine Zaoui / @nzaoui : The term deepfakes comes from the Reddit username of the person or persons who in 2017 released a series of pornographic clips modified using machine learning to include the faces of Hollywood actresses. https://www.wired.com/... @lishali88 : .@tsimonite 's nuanced coverage of @Rosebud_AI's synthetic model imagery being used by brands during Covid to make marketing photo production possible while SIP. https://www.wired.com/... @wired https://twitter.com/... https://twitter.com/... Fareed Mosavat / @far33d : Great Wired article on @Rosebud_AI and other synthetic imagery tools that enable a new use cases that were impossible or costly before. This could have as big an impact as 3D graphics and visual effects - augmenting vs. replacing live action shoots. https://www.wired.com/... Chris Messina / @chrismessina : Some solid coverage of #SyntheticMedia in the time of #COVID19 ... if models and photogs can no longer get together IRL, why not just fabricate realistic content using AI and ML? /cc @MattHartman @Borthwick #DeepFakes #Advertising #SyntheticReality https://twitter.com/... @nxthompson : Expired: deepfakes for porn Tired: deepfakes for political chicanery Wired: deepfakes for corporate advertising. https://www.wired.com/...
Context & Ripple Effects
Deepfake tech began as a Reddit-born pornography phenomenon and stayed there for years — one author demonstrated how accessible it had become by producing a convincing fake over two weeks for just $552 using open-source Faceswap. What changes with this Wired piece is the customer: Synthesia is steering the same underlying technology at corporate clients, generating multilingual and personalized training videos instead of celebrity fakes.
The pivot is already drawing followers and friction. A year later, Tel Aviv's Hour One formalized the model by paying people to license their likenesses for AI-voiced marketing and educational videos, while Rosebud AI's synthetic fashion models — used by brands during COVID — prompted an AI professor to warn that diverse synthetic humans could end up substituting for real people from minority communities.
First-order effects
- Synthesia hands corporate training and L&D teams a sanctioned way to buy video at scale — one recorded presenter becomes many languages and personalized variants without reshoots.
- A paid-likeness market opens up: Hour One's follow-on model shows performers can be compensated for synthetic use of their face and voice, establishing consent as the transaction.
Second-order effects
- Rosebud AI's brand work forces modeling agencies and marketers to confront synthetic substitutes — and the specific warning that 'diverse' avatars may displace real talent from underrepresented groups puts pressure on buyers' casting choices.
- As legitimate uses normalize the tools, the same accessibility documented in the deepfake meme tutorials spreading on YouTube lowers barriers outside the enterprise context, widening the gap between licensed and unlicensed use.
Third-order effects
- Synthetic media splits into two economies: a consent-based B2B tier (training, marketing, education) and the nonconsensual tier already dominant in usage — where, as coverage of the porn-driven deepfake ecosystem shows, stigma itself blocks victims' IP protection, making provenance and verification infrastructure the structural battleground.
- If enterprises become the anchor customers for generative video, vendor selection will hinge on governance — audit trails for who approved which likeness — pushing the industry toward certification rather than raw capability.
The trend: Deepfake technology is bifurcating from its nonconsensual origins into a licensed, enterprise-grade synthetic media business built on paid likeness agreements.