How To Make AI Avatars Talk - A Step by Step Guide

How To Make AI Avatars Talk - A Step by Step Guide

Recent AI developments keep blowing my mind! In my previous article Lensa AI Can Turn Your Selfies into Magical Avatars, I introduced the app Lensa AI that can transform selfies into magical avatars like a warrior, king or queen, cartoon character and much more. Recently I discovered a new cool AI tool that can make animated AI-generated characters talk. Read on and check out the YouTube video below!

I made the AI avatar narrate If by Rudyard Kipling. It turned out surprisingly well! The mouth of the AI avatar moves naturally. The voice sounds real and has a lot of passion.

How can you make an AI avatar talk like the videos above?

The three YouTube videos that you watched above were made using a combination of tools:

  1. The AI avatar is using AI image generation software like Lexica, Mid-journey and Lensa AI.
  2. The video animation is using D-ID, a tool that can generate AI avatar videos
  3. The voice is generated using ElevenLabs, an AI speech synthesis software

Below is a step-by-step guide.

Step 1: Pick an avatar and uploaded it in D-ID studio.

The avatar can be an AI generated avatar or a photo of a real person.

  • If you'd like to use an AI avatar, you can generate one using AI image generation software like Lexica, Mid-journey and Lensa AI. You also have the option of writing a prompt to generate an avatar within the D-ID software.
  • If you don't want to use an AI avatar, you can upload a photo of the real you (or someone else if that person does not mind).
  • I tried to see if uploading pictures of animals would work because I wanted to create a video of a panda talking :) Unfortunately it didn't work. It looks like the software can only recognize real human faces or AI faces, not animals.

Step 2: Decide what voice to use. There are 2 choices here:

  1. Upload an audio. You can either upload a real human voice or using AI synthesized voice. In the 1st video, the voice is generated using ElevenLabs, an AI speech synthesis software. I find the voice very real and passionate. By default, the speed in ElevenLabs can be a bit fast sometimes, but a useful tip is to dd punctuation (for example "...") to slow down the pace of the voice.
  2. Type your script, choose your language and voice style within D-ID. There are several choices of different languages and voice styles you can choose. In the 2nd video, the voice is picked from the D-ID voice styles.

Step 3: Click "GENERATE VIDEO". Done!

Please see screenshot below for an illustration of the steps.

No alt text provided for this image

I was so blown away by how naturally the AI avatar could talk. The videos turned out surprising well. Let me know what you think after you create your own videos with talking avatars!

Benjamin Sebagh

Head of IA @Jaws Group | Formateur & Consultant en IA Générative

9 个月

This works with human pictures, what about having a dog picture to make it talk with mouse movements ?? Did you saw any AI for that ?

回复
Richard Djarbeng

Web, data and Internet of Things (IoT) | Environmental monitoring | AI

1 年

Beautiful. Did you perhaps post this on twitter as well?

Michael Doyle

Author: Future-Proofing Start-ups at i3D Protocol by Invluencer

1 年

I've used the software quite extensively to create knowledge videos. Very easy to use. Unfortunately the pricing model is quite expensive as soon as you want to ramp up into production mode. In this competitive space they'll have to revisit this as more competition enters.

要查看或添加评论,请登录

社区洞察

其他会员也浏览了