Vantaige

System Diagram

Inspiration

I am a solopreneur and even though I have built a team of developers and a business developer, we ran into the simple issue of "How do we market this? Where do we even start?" We don't have any capital or connections with anyone who will do marketing and design for us at a the budget that we are at so when I saw this Gemini Live challenge, I was intrigued if we can actually build something that can offset our competence gap.

What it does

Vantaige is a marketing director with conversation as the main mode of use. You simply start a session and speak with vantaige. Vantaige directs the conversation naturally towards marketing as it helps teams without prior experience with marketing create ready to post instagram and facebook posts with hashtags included, as well as generate tiktok and youtube shorts. Vantaige does this with the directive of keeping the brand consistent across platforms.

How we built it

We build this using Gemini Live API through vertex. The requirements of using google cloud inspired us to fully utilize google cloud products:

Google Vertex for Gemini Live, Imagen, and Veo access.
Firestore for NoSQL database.
Firebase storage for storing generated assets.
Google Cloud Run for app hosting

For the application stack, we used FastAPI for the backend and Nextjs for the frontend.

Challenges we ran into

One of the challenges we ran into is finding app examples that works. Unfortunately, across multiple releases, there's quite a few competing documentation on how to use Gemini Live API. We also had trouble accessing certain models as they're on preview and had to settle for a slightly older model instead. Veo 3.1, as powerful as it is, still has trouble with frames with text on it. The Live API also had issues sometimes with timing or competing instructions but they were easily fixable by adjusting the prompts.

Accomplishments that we're proud of

I am very proud of how well the marketing flow works across sessions and the generated images for facebook and instagram posts are, for most of the time, ready to post. I am also very proud of how well the conversation as the main mode of control for the app works. For most of the sessions, I barely even used my mouse and keyboard. As I was testing this, I realized how accessible the app became, specially for people with disabilities where conversational control of the application could help a lot.

What we learned

First of all, I learned a lot about multimodal models and how easy it is to use. I was particularly impressed by how the live api feels so seamless. Secondly, I learned how to leverage GCP for my future projects. Having everything hosted on google cloud really helped with the deployment simplicity. Lastly, using vantaige actually helped me thing in terms of marketing and how I should go about it. This is absolutely something that I was aiming to get as well. Using Gemini Live API helped me process information way faster.

What's next for Vantaige

The natural next step for vantaige, unsurprisingly, is to use vantaige to market vantaige. As I built vantaige, I can see the utility of the product with a specific market segment targeted.

In terms of functionality, I want to add direct integration to google ads, and other ad spaces so the user never has to leave the application. Another integration would be adding an MCP and API connection to pull information into other platforms.

It would also be great if we can monetize vantaige as the cloud bill will surely stack up given the high usage of Veo and Imagen