How to turn an image into a talking video
Turn a single still image into a lip-synced talking video with sync-3. No source footage, no avatar setup: one photo in, a speaking person out.
When people want to learn about a product, 63% say they’d rather watch a short video than read about it, against just 12% who prefer a text article, per Wyzowl’s 2026 State of Video Marketing report. A still image holds attention for a moment; a person talking holds it for the message.
sync. labs turns a single still image of a person into a talking video. You upload one photo, give it audio, and sync-3 builds the performance, the mouth, the face, the head, all the motion that makes a still feel alive. No source footage, no avatar to train, no green screen. One frame in, a speaking person out, usually in under three minutes.
That used to be impossible. Lip sync needed video of someone already talking, because older models only edited motion that was already there. sync-3 generates the motion instead of editing it, which is why a single image is now enough.
Image-to-video builds a performance from one frame
sync-3 reads the whole face in a still and generates every frame of the talking video at once, instead of patching a small region around the mouth. That global read is why it can start from a single image: a full video and one frame are just different amounts of the same information, and sync-3 can work from the smallest amount.
This isn’t a separate tool bolted onto the side of the product. It’s the same model that powers lip sync on real footage, asked to start from less. Image input runs on sync-3 because nothing else understands a face well enough to build one a frame at a time.
Any clear still of a person can talk
The unlock is that your input no longer has to be video. A headshot, a product spokesperson, a historical portrait, a character illustration, an AI-generated face: if it reads as a person, sync-3 can make it speak. A clear, front-facing still with the face well lit and unobstructed gives the cleanest result, but the model holds up across a wide range of inputs.
You can turn an image into a talking video in four steps
- Upload a still image. One clear photo of a person. png and jpg both work, no conversion needed.
- Add the audio. Type a script and pick a voice, upload an audio file, or use a clone of your own voice so the image speaks in it.
- Generate. sync-3 builds the talking video from the frame, lip synced and expressive, in one pass.
- Export or translate. Render at the resolution you need, and generate the same talking video in any of 95+ languages from the same image.
The whole flow runs in the browser, no editing suite required, and the first three videos a month are free.
A talking video carries a message a static image can’t
| Static image | Talking video from the image | |
|---|---|---|
| Delivers spoken words | No | Yes |
| Holds attention | Briefly | Through the message |
| Works across languages | Caption only | Spoken in 95+ languages |
| Source needed | One photo | The same one photo |
The point isn’t that images are dead, it’s that the same photo can now do more. The asset you already have becomes a presenter, a narrator, or a spokesperson without a shoot.
How sync. labs compares to other image-to-video tools
Several tools turn a photo into a talking video. They differ on whether the result speaks in your own voice and how much setup it takes.
| Tool | From a single still | Voice | Languages |
|---|---|---|---|
| sync. labs | Yes, on sync-3 | Your cloned voice, uploaded audio, or generated | 95+ |
| HeyGen (Avatar IV) | Yes | Cloned, recorded, or text-to-speech | 175+ |
| D-ID (Creative Reality) | Yes | Cloned or stock voices | 40+ |
| Hedra (talking avatar) | Yes | Cloned or 4,000+ stock voices | Dozens |
| Synthesia (personal avatar) | Yes, needs a consent recording | Cloned or stock voices | 160+ |
| Runway (Add Dialogue) | Yes | Synthetic text-to-speech only | Varies |
Details are from each tool’s own site as of June 2026: HeyGen, D-ID, Hedra, Synthesia, Runway. The difference that matters is realism and voice. sync-3 generates the full performance from one frame on the same model that does studio lip sync on real footage, holds up on hard shots, runs up to 4K at 60fps, and speaks in your own cloned voice rather than defaulting to a synthetic one. It’s available in a web playground, a REST API, an Adobe Premiere plugin, and a ComfyUI node.
Frequently asked questions
Do I need video footage to make a talking video?
No. sync-3 builds the talking video from a single still image. You don't need any existing video of the person, just one clear photo and audio.
Whose voice does the image speak in?
Your choice. You can clone a voice so the image speaks in it, upload an audio file, or generate a voice from a typed script. Cloning keeps the real pitch, tone, and cadence.
What kind of image works best?
A clear, front-facing still of a person with the face well lit and unobstructed gives the cleanest result. Headshots, portraits, character illustrations, and AI-generated faces all work.
Can the talking video be in another language?
Yes. The same image can speak in 95+ languages, so one photo becomes a talking video in Spanish, Hindi, Japanese, French, and more.
Can I turn an image into a talking video for free?
Yes. sync. labs includes three free videos a month at full HD, with no credit card required. Higher volume moves to a paid plan.
Turn your first image into a talking video on sync-3. For how the model builds a performance from one frame, read the image-to-video launch.