A Talking AI Avatar From One Photo and Your Own Voice: The Complete Guide

One photo of you and a voice, and every new text becomes a video of you talking to camera. Which photo works, how to clone your own voice, and how to end up with a video that looks natural.

A talking avatar is a video in which a character looks into the camera and says a text you wrote, with lips that move to the words, facial expression and natural head movement. Until recently that took a studio shoot, lighting, and a new recording every time a sentence changed. Today one photo of you and a voice are enough, and every new text becomes a new video without standing in front of a camera again. This guide explains what you need, which photo works best, how to clone your own voice, and how to come out of it with a video that looks natural.

Who It Is For

Business owners who want to explain a product every week without filming every week. Creators who post daily and do not always have the energy to be on camera. Teachers and lecturers who record lessons. And anyone who needs the same video in Hebrew and in English, with the same face and the same voice.

What You Need: One Photo and a Voice

• A photo of you, or a character you made with AI
• A voice: a ready voice from the library, or a clone of your own voice from a short recording
• The text the avatar will say, written in the language it will speak

You build the avatar once. From then on it stays yours, and for every next video you only write a new text.

The Photo That Works Best

• A close portrait: head and shoulders, with the face filling most of the frame
• Eyes straight into the camera, not from the side or from below
• Even light on the face, with no hard shadow across half of it
• Nothing on the face: no sunglasses, hat, mask or hand
• Mouth closed or in a relaxed smile. A wide-open mouth looks odd once it starts talking

The Voice: Ready-Made or Cloned

There are two ways. The first is to pick a ready voice from a voice library, in Hebrew, English and many other languages. The second is to clone your own voice: you record yourself talking, and the system builds a voice from it that reads any text you write, even when you never recorded it. A clone works from about 10 seconds of clean speech, but a minute or two of ordinary speech gives a voice people recognise. The cloned voice speaks more than 90 languages, Hebrew included, so the same voice of yours can explain the product to customers abroad as well.

• A quiet room: no loud air conditioner, TV or music behind you
• Speak the way you want the avatar to speak: the same energy, the same pace
• Mix in a question, some excitement and a quiet moment, so the voice learns your whole range
• No long silences at the start or the end

Worth knowing: Clone only a voice that is yours, or one whose owner gave you permission. The same rule applies to a photo of a real person.

Step by Step

• Build an avatar from the photo. It happens once, and it stays on your shelf
• Pick a voice from the library, or clone your own voice from a recording
• Write the text in the language the avatar will speak
• Pick a format: vertical 9:16 for Reels or wide 16:9, in 1080p or 720p, with a colour or picture background
• See the price, approve it, and get the video with a synced caption file for free

Build an avatar from the attached photo, attach my voice to it, and have it say, vertical: Hi, welcome to the studio

More Than a Talking Head

• The same face in other clothes or another room, or inside a cinematic shot you describe in words
• Up to 4 avatars in one film, each in its own voice and language, and you mark who says each sentence
• A whole film from a text or an idea: shots, pictures, music and captions around the avatar
• Translating a finished video into 8 languages, with the lips moving to the new language
• All of it can also be asked for from inside ChatGPT and Claude, with the same avatars and the same voices

What It Costs

You build the avatar once, and after that each video is priced by the length of the speech. The price appears before anything runs, and nothing starts without your approval. If a creation fails for a technical reason, the credits return to your balance automatically. There is no monthly subscription: you buy one credit package, and it works for avatars, voices and every other engine.

Mistakes That Give It Away

• A text in one language with a voice set to another. The avatar says exactly what is written, so write in the language it will speak
• A photo with an open mouth, sunglasses or a hand on the chin
• A cloning recording with background noise, which is then heard in every sentence
• A text too long for one video. Two short ones work better
• Sentences written to be read rather than spoken. Write the way you talk

Questions and Answers

Can the avatar speak Hebrew? Yes. Both a ready voice and your cloned voice speak Hebrew, and the same avatar can speak Hebrew in one video and English in the next.

Can I use a character who does not exist? Yes. You can build an avatar from a character you made with AI, so the business has a permanent presenter without anyone being filmed.

Does the video belong to me? Yes. What is made in your account belongs to you, for personal, business or commercial use.

Start Now

In the Cadabra AI Avatar Studio you build an avatar from a photo, attach a voice and get a first video in the same visit. A new account opens with welcome credits. The full guide, with screenshots of every step, is on the "Avatar Studio guide" page on the site.

Where to go next

• 105 image codes, and here they run
• 61 codes that turn a picture into a clip
• More guides on the blog