Then I came across Openspline, whose tagline is "Give your AI a face." I was skeptical. Avatar tools usually mean heavy setup, expensive plans, or being locked into one platform. But a few things made me try it:

• It plugs into stacks I already use: ElevenLabs, LiveKit, OpenAI Realtime, Gemini Live, Pipecat, or any audio stream
• I could either self-host it or use their cloud

What it actually does

The idea is simple: you keep your own model, prompts, and voice. You send audio in and get synchronized video back.

Openspline doesn't replace your agent. It's the face that sits on top of it. That mattered to me because I didn't have to rebuild anything, and my ElevenLabs voice stayed exactly the same.

Setting it up

There are two ways to run it, and I went with the cloud.

Self-host: you need an NVIDIA GPU and Linux or WSL2. The install is a one-liner from their docs.

Openspline Cloud: you create an avatar in the browser with a face, a voice, and a personality, then embed it. There's no GPU to manage.

The code

This was the part that surprised me. Here's the core of the Python SDK, straight from their docs:

from openspline import Openspline

client = Openspline()

async with client.avatar("portrait.jpg") as avatar:
    print(avatar.viewer_url)
    await avatar.stream(
        agent.audio_stream(), sample_rate=24000
    )

You give it a portrait image, open an avatar session, and stream your agent's audio into it. You get back a viewer_url where the face appears and talks.

Where I'd use this

Openspline's site suggests customer support, AI tutors, interviews, and sales. I used that for fun making Leonardo Da Vinci to roast my startup idea.

• Heres the link : Link

If you already have a voice agent and want it to feel less like a phone call, give it a try:

• Docs: docs.openspline.dev
• GitHub: github.com/openspline/openspline
• Website: openspline.dev

If you try it, drop a comment and tell me what you build!