Video creation time reduced from days to minutes
Gemini 3 unlocked our Motion Graphics engine - success rate improved from <10% to high 90s
Step change in visual intelligence with Gemini 3
HeyGen uses Gemini to power its Video Agent, reducing video creation from days to minutes and delivering professional quality videos.
HeyGen uses Gemini to power its Video Agent, reducing video creation from days to minutes and delivering professional quality videos.
Making video storytelling accessible to all
Our Video Agent requires visual and aesthetic intelligence, advanced reasoning, long context management and more. We evaluated pretty much all the LLM models in the market and Gemini 3's superior multi-modal intelligence puts it way ahead and an easy choice for us.
Bin Liu
VP, Product Engineering, HeyGen
Creating professional video has always been time-intensive and expensive, requiring technical expertise and comfort in front of a camera. For businesses creating training materials or marketing content, this barrier can mean that valuable knowledge stays locked in documents nobody reads.
HeyGen set out to change this. With just one uploaded photo or video, HeyGen enables customers to create studio-quality videos that educate, inform, and sell— from product training videos to bespoke customer pitches. With millions of users and approximately one million videos created daily, HeyGen has made video production accessible to all.
HeyGen recently launched AI Video Agent (available through web app and API), raising the bar by enabling users to generate polished videos from simple prompts.
From creating social ads to product explainers, the agent performs tasks a professional video team would handle: writing scripts, selecting visuals, adding voiceovers, and applying cuts, pacing, and transitions.
Building Video Agent required an LLM capable of analyzing, structuring, optimizing, and processing video content with human-level understanding. "As we've evolved from a platform where users come to HeyGen to create a Digital Twin of themselves to a more holistic platform where users can also take that footage and use an agent to edit it directly, we needed a lot more intelligence in the AI technology we use," says Bin Liu, VP of product engineering at HeyGen. "We tested pretty much all the LLM models in the market and chose Gemini for superior performance."
Driven primarily by the rapid adoption of Video Agent, HeyGen has seen 10x growth in Google Cloud spend since the first quarter of 2025. "The level of quality we can rely on with Gemini to build our products means we can sleep at night," says Liu. "We don't have to worry about it."
Building intelligent video creation with Gemini
HeyGen, which joined Google Cloud for Startups when the company was founded in 2020, runs approximately 80% of its technology on Google Cloud and leverages Gemini through Vertex AI.
"Out of all LLM providers, Gemini's the only one that truly has visual intelligence with a sense of aesthetics," says Liu. "This meant our Video Agent immediately got so much more capable with Gemini, enabling so many things our platform otherwise wouldn't be able to offer to our users."
Since switching over to Gemini 3, the quality of content, pace, aesthetics all reached new heights based on our internal side-by-side evals. Accuracy improved from the low 10s to the high 90s with the adoption of Gemini 3.
Our Motion Graphics engine's quality and success rate has more than 10x-ed with Gemini 3. Professional business users come to us because they value simplicity and quality, and this level of multi-modal intelligence massively uplevels their visual capabilities.
Bin Liu
VP, Product Engineering, HeyGen
HeyGen has piloted Gemini 3 before a full rollout, and Liu says the latest version is a significant leap forward. "Our internal prototype became a production pipeline overnight with Gemini 3's launch," says Liu. "Our customers come to us because they value simplicity and convenience, and this level of intelligence massively uplevels their visual capabilities. It's time-saving for our users, cost-saving for us, and the quality went up multiple levels thanks to Gemini 3's visual intelligence."
HeyGen's use of Nano Banana Pro and Veo 3 has solved a persistent video production challenge. Previously, users would use their actual background or change to a virtual background. Now, Video Agent helps them accurately mimic their avatars in any environment. Additionally, HeyGen's Video Translate uses Gemini 2.5 to translate full video scripts into over 180 languages, while its Proofread tool refines translations and allows customers to adjust voices and scripts. "Gemini has helped us to deliver truly global, personalized experiences," says Liu.
Reinventing the future of video
We are leaning into Gemini 3 because it's a real goldmine. It allows us to reinvent what video creation and editing can be.
Bin Liu
VP, Product Engineering, HeyGen
HeyGen also uses Gemini to power a comprehensive evaluation system. It runs extensive internal offline evaluations to validate video output quality, relying on Gemini to assess whether videos follow prompts accurately, use the correct avatars, maintain good pacing, and engage viewers in the critical first 10 seconds. "We ask Gemini specific questions to evaluate across all of our videos how well our system is doing, and this is critical as we evolve," says Liu.
The team also conducts qualitative side-by-side comparisons to determine whether new LLM models meet HeyGen's performance thresholds, set by Gemini. This rigorous, practical approach to evaluation ensures that every improvement translates to tangible benefits for users. "We have a bunch of scores that LLMs need to beat, and if they don't, then we simply won't use them," says Liu. "Honestly, you'd be surprised how many LLMs can't deliver on that."
Looking ahead, HeyGen plans to test more Google Cloud AI products, including future versions of Veo and Nano Banana, as well as the music generation model, Lyria. Fundamentally, though, HeyGen is focused on deepening its use of Gemini 3 in Video Agent. "We are leaning into Gemini 3 because it's a real goldmine," says Liu. "Gemini 3—and the suite of state-of-the-art APIs and AI models from Google Cloud—allow us to reinvent what video creation can be."
HeyGen is transforming storytelling with AI, empowering anyone to create high-quality videos without limits.
Industries: Technology, Startup
Location: United States
Products: Google Cloud, Gemini, Lyria, Nano Banana Pro, Veo 3, Vertex AI
