How to make an AI avatar with your face and voice for Reels
Record a 3–5 minute video and start training: 2.5–3.5 hours and 500 credits (≈ $12), once. After that the avatar speaks in your voice.
To create an AI avatar with your face and voice, record a 3–5 minute video, upload it to Craft and start training. Training takes 2.5–3.5 hours and costs 500 credits, about $12, once. After that a Reels is built in 5–10 minutes from your own text or from someone else's viral video.
Examples of a real person next to their avatar are on the Content Factory page. All the settings are covered in detail in Avatar reels.
What you'll need#
- A video of you talking to the camera. The recording rules are below.
- A Craft account. Signing up is free.
- 500 credits for training, plus credits for videos.
⚠️ Your own face, or a face with consent. You can upload only your own face, or the face of a person who has agreed in writing to an AI avatar.
Step 1. Record a video#
What affects the result most is the source video: lighting, a close-up face, silence.
- 3–5 minutes of speech, calm, like a regular Reels. We won't accept less than 15 seconds.
- The camera stays still and you sit. Your face is in frame the whole time, close up, from the chest up.
- Your face is well lit, and your mouth isn't covered by a hand or a microphone.
- Quiet around you. Music and other voices end up in the voice clone.
- At the start and in pauses, stay silent for 2–3 seconds with your lips closed.
- Keep gestures below your face. A repeated gesture becomes the avatar's habit.
Step 2. Upload the video#
Open Content Factory → Avatar reels and upload the file. Uploading and checking are free. Craft checks the video within a minute and, if something's wrong, says what on the card. Then it makes a sample phrase in your voice: press the Hear the voice player.
If the file is over 400 MB, upload it to Google Drive, Yandex Disk or Dropbox, share it with "anyone with the link" and paste the link under the cards. Craft downloads files up to 8 GB itself, compresses them to 1080p and takes the first 5 minutes.
Step 3. Start training#
On the avatar card, press Train avatar. The price in credits is shown next to it. You can close the page, and the card shows the stage and percentage. Credits are charged at launch, and if training fails, they come back automatically.
Training is done once per avatar. You only need to retrain if you upload a new video.
Step 4. Check the likeness#
You'll judge for yourself before the first video. After training, a preview appears on the card: your phrase in your voice on your face.
- Lips. Open the avatar window and watch the tests under Training depth. The avatar is trained at three depths, each with its own test. Press Choose where the lips move most naturally.
- Voice. Listen to the sample. If you don't like it, press Choose voice under the player. There are several voices, each with its own test phrase. After training, "My voice" appears, and it sounds livelier.
- Face. Under Reel quality, pick Standard, High or Maximum, make a test for a few credits and compare.
Step 5. Make your first video#
Give a link to someone else's Reels, your own text or just a topic. Craft writes a script in your brand's voice, and you edit and approve it. Nothing is charged before you approve. When the video is ready, you'll get a notification in the bell and in Telegram.
How long it takes#
| What | How long |
|---|---|
| Video check | About a minute |
| Avatar training | 2.5–3.5 hours, once |
| Video with the avatar | 5–10 minutes, up to 20 for the first one after a break |
| Voiceover and editing without a face | 1–3 minutes |
How much it costs#
| What | Credits | Approx. |
|---|---|---|
| Signing up, uploading and checking the video | 0 | $0 |
| Avatar training, once | 500 | $12 |
| Your very first 40-second video | 174 | ≈ $4.05 |
| A 40-second video from the second on, for good | 160 | ≈ $3.73 |
Dollar amounts are approximate, at ≈ $0.023 per credit. With volume the price drops to ≈ $1.40 per video: the calculation is in Content factory cost. A pack of 1,500 credits for $35 is a good start: enough for training and about 6 videos of 40 seconds. Purchased credits never expire, but they can't be refunded as money, so start with a small pack.
Common mistakes#
| What you'll see | Why | How to fix it |
|---|---|---|
| Training failed, the card shows a reason | Your face is barely visible in the video: you turn away, the face is small or leaves the frame | Reshoot following step 1. The credits come back automatically |
| The lips don't match, the face distorts | Most often the culprits are lighting, a small face and a hand near the face | Reshoot the source and train again |
| You hear noise or someone else's speech in the voice | Music, echo and other voices got into the clone | Record the video in silence or pick another voice |
| The avatar keeps repeating one gesture | The gesture was repeated a lot in the recording | Record a new video and keep gestures below your face |
| The message "The avatar isn't trained yet" | Training wasn't started | Start it on the avatar card |
| The message "Not enough credits" | Your balance is below the price of the video | Top up your balance, and the video will start after that |
| A voice sample is bad or wasn't voiced | It happens with individual samples | Press Voice again next to that voice |
What's not there yet#
- A completely free avatar. Training costs 500 credits.
- Equal quality in every language. The avatar works best in Russian. Videos in English, Chinese, Hindi and Arabic have already come out. In other languages you can hear an accent in the voice, and we haven't checked lip sync there yet. You can connect your own ElevenLabs key and use a voice from there.
- Guarantees of views and sales. What takes off is up to the platform and your topic.
- Customer case studies with numbers. The first customers are making their first videos, and we'll show numbers once they're real.
FAQ#
Can I create an AI avatar for free?#
Signing up is free, and so are uploading and checking the video. Avatar training costs 500 credits, about $12, once. If it fails, the credits come back automatically. Without training, "Voiceover only" and "Editing only" are available, and they're much cheaper.
What video do I need to record for an AI avatar?#
3–5 minutes of talking to the camera, 15 seconds at the very least. The camera stays still, you sit, your face is in close-up and well lit, and your mouth isn't covered by a hand or a microphone. Keep it quiet around you, because other voices and music end up in the voice clone.
How long does it take to create an AI avatar?#
Training takes 2.5–3.5 hours, once. You can close the page. After that a video takes 5–10 minutes, up to 20 for the first one after a break.
Will the AI avatar look like me?#
You'll judge for yourself before the first video. After training, a preview in your voice on your face appears on the card. The avatar is trained at three depths, and for your videos you pick the one where the lips move most naturally.
Can I make an AI avatar of someone else?#
Only if they have agreed to an AI avatar in writing. You can't upload someone else's face without their consent.
Which languages can the AI avatar speak?#
You pick the language when you create a video: there are 44 on the list. The avatar works best in Russian. In other languages you can hear an accent in the voice, and we haven't checked lip sync yet.