Creating voice Agent was never easier. In past a lot of provider were charging heavy. And testing the AI models with the audio was not that easy in the past. Now a days a lot of providers are offering free audio models.
Take example of Fish audio which is offering their S2 model. And you can try out the developer API for free, check out Fish audio API.
I created an Audio agent and the voice model S2 is being used in one of the agent I created. This is a shelter agent which makes it easy to use in the existing website or mobile app. You can check out the video below to see how this is being done.
How this was made?
First thing I did was created a dummy data. Which made it possible to for me to have an agent build over. For example if you connect the Vercel's Eve framework. This would require something on which the agent to be used.
Once you create dummy data and sample questions the agent would be answering. You have given a head start. And from that onwards you can ask the agent for the voice agent questions from the sample list.
Use the fish audio API and choose the female or male voice on which you want to train the agent. Once you do that you are good to go. You can even explain your process verbally and most of the AI agents would be able to build the rest.
I used Antigravity IDE and asked it to build this agent. Here I explained about dummy data. And also explained about openrouter's free models. And got Vercel's Eve framework to build the chat agent interface.
That's it. And you have the final output as explained in the video above. Make sure to experiment with your dummy data. You can even take the voice input for ai agent to listen to and then get the output in the voice in return too.
Top comments (0)