Make an Old Photo Talk With Your Voice Using AI

To make a photo talk with your voice, record what you want said, then upload the photo and the audio file to VdoBloom’s AI Talking Avatar; the AI animates the face so the lips move with your recording. For an old photo of a grandparent, restore it first in Photo Restore (5 credits) so the face is sharp. A talking clip is priced per second of audio: InfiniTalk at 480p is 1.2 credits a second with a 15-credit minimum, and OmniHuman 2.0 is 5.6 credits a second for up to 5 minutes. New accounts get 10 free credits, enough to restore two photos. The tool does not clone voices: it speaks with the audio you upload.

Quick facts

• What you need: one photo with a clear face, and an audio file of the words.
• Voice: your own recording, or a voiceover from Text to Speech. No voice cloning of other people.
• Restore first: Photo Restore fixes scratches, creases, fading and blur for 5 credits, and keeps black and white unless you pick Colorize.
• Talking models: InfiniTalk (up to 120 s), OmniHuman 1.5 (up to 60 s), OmniHuman 2.0 (up to 300 s, full-body gestures and prompt direction), plus Kling AI Avatar (up to 5 minutes).
• Cheapest 10-second test: 15 credits on InfiniTalk 480p.
• Free start: 10 free credits on signup; one-time packs never expire, no subscription needed.
• Consent: only use photos and voices you have the right to use. Be careful with memorials (see below).

How do I make a photo talk with my voice?

1. Record your voice. Use your phone’s voice recorder in a quiet room, about a hand’s width from your mouth. Speak at a natural pace. Save or export it as an audio file such as MP3, M4A or WAV.
2. Open the AI Talking Avatar and upload the photo.
3. Upload the audio. The tool reads its length, so the price shown matches your clip.
4. Pick a model (see the table below). For a first test, InfiniTalk at 480p is the cheapest.
5. Generate. The face is animated to your audio and the video is saved to My Creations.

Prefer not to use your own voice? Type the words into Text to Speech (MiniMax Speech 2.8, 26 voices, 40 languages including Hindi and Tamil, 8 credits per 1,000 characters), download the MP3 and upload that instead. Or describe a voice in AI Voice Design.

How to make an old photo of grandpa talk with AI

Old prints need one extra step, because the talking model can only animate detail it can see.

1. Scan or photograph the print

A flatbed scan is best. With a phone, shoot straight on in even daylight, with no glare, and crop out the album page.

2. Restore the photo

Upload it to Photo Restore. Choose Light wear for fading and fine scratches, or Heavy damage to also fill small tears. Keep the original tone, or pick Colorize for natural, period-accurate colours. Restoration costs 5 credits and is usually ready in under two minutes. It only sharpens detail that is already there, so grandpa’s face, age and expression stay his.

3. Crop to head and shoulders

If grandpa is one face in a group photo, crop so his face fills a good part of the frame. A tiny face in a crowd gives the model too little to work with.

4. Add the audio and animate

Upload the restored photo and your audio to the talking avatar. Many families use a recording of a relative reading something grandpa used to say, a poem, or a birthday message. You can also use an old recording of his real voice, if you have one and the family agrees.

Just want gentle movement, no speech?

Use Animate Old Photos instead. It turns a still into a short, soft memory video (a smile, a blink, a turn of the head) without lip-sync.

Which talking model should I pick?

Model	Max audio	Credits	10-second clip	Good for
InfiniTalk, 480p	120 s	1.2 per second, 15 minimum	15	Cheap first test
InfiniTalk, 720p	120 s	4.8 per second, 15 minimum	48	Sharper version of the same look
OmniHuman 2.0	300 s	5.6 per second (rounded up)	56	Natural head and body motion, longer messages
OmniHuman 1.5	60 s	10.8 per second	108	High-quality lip-sync for short clips

A good habit: test the first 10 seconds on InfiniTalk 480p, and once the photo and audio work, render the full message on OmniHuman 2.0. The models available in the picker can change; the price for your exact clip is shown before you generate.

What does it cost in total?

Project	Steps	Credits
Quick test	InfiniTalk 480p, 10 s	15
Grandpa’s 20-second birthday message	Restore (5) + OmniHuman 2.0, 20 s (112)	117
A 1-minute memorial message	Restore (5) + OmniHuman 2.0, 60 s (336)	341

See the pricing page for packs and plans.

Tips for natural lip-sync

• Front-facing photos work best. A three-quarter view is fine; a full profile is not.
• Mouth visible and closed or relaxed in the photo. A big open laugh or a hand over the mouth makes the lips look odd.
• Clean audio. Background music, echo or a TV in the room confuses the lip-sync. Record in a small, soft room.
• Short sentences, natural pauses. The face looks calmer when the speech breathes.
• Keep it short first. Per-second pricing means a 10-second test costs little, so check the look before you render a full minute.
• Restore before you animate. Scratches across the mouth or eyes move with the face and look worse in video.

Consent and respect for family memorials

A talking photo of someone who has died can be comforting, and it can also be upsetting for relatives who were not asked. A few simple rules:

• Talk to close family first, especially a surviving partner or the person’s children.
• Put words in their mouth only if they are words the person really said or would be comfortable with. Avoid jokes or messages they would not have wanted.
• Say clearly that it is an AI animation when you share it, so nobody mistakes it for a real recording.
• Do not clone or imitate someone else’s voice without their consent. Use your own voice, a family member who agreed, or a stock voice.
• Never use this to make a living person appear to say something they did not say.

Frequently asked questions

How do I make a photo talk with my voice?

Record your voice as an audio file, then upload the photo and the audio to the AI Talking Avatar and generate. The lips are synced to your recording. A 10-second clip starts at 15 credits on InfiniTalk 480p.

How to make an old photo of grandpa talk with AI?

Scan the print, restore it in Photo Restore (5 credits), crop to his head and shoulders, then upload it with an audio recording to the AI Talking Avatar. Talk to the family first and label the result as an AI animation.

Can the AI copy grandpa’s real voice?

No. The avatar does not clone voices. The avatar speaks with the audio you upload, so you can use a real old recording of him if you have one and the family agrees.

How long can the talking video be?

Up to 300 seconds of audio on OmniHuman 2.0, 120 seconds on InfiniTalk and 60 seconds on OmniHuman 1.5. Kling AI Avatar takes up to 5 minutes.

Does it work with black-and-white photos?

Yes. Restore keeps black and white unless you choose Colorize, and the talking avatar animates either.

Is it free?

New accounts get 10 free credits, which covers two restorations. The talking video needs at least 15 credits, so a small top-up is needed for the first clip. One-time pack credits never expire.