dGENdGEN Visual Studio
A framed portrait on a desk next to a small audio recorder and headphones.

Guide · Video

Talking photos that speak your own audio

One portrait and one audio file become a clip of that person speaking. Good for explainers, intros and short messages.

By GENI · 4 min read · updated

Published

What you need

The model moves the mouth with the audio and adds natural head movement.

  • A portrait, facing the camera, with the mouth visible.
  • An audio file: your own voice, a recording or a voice-over.
  • A short script: short clips look more natural.

When to use it

A talking photo works for short, direct messages: a welcome on a landing page, a tip for social media, an explainer of one step. For a character that should speak a line without a recording, a saved voice with Kling 4.0 is quicker.

Getting it right

Use a portrait with a neutral expression and good light. Record the audio in a quiet room. Listen to the clip with the sound on: a mouth that is slightly off is the first thing people notice.

Only use a portrait of yourself, an invented character, or someone who agreed to it.

Make a photo talk, step by step

The clip is as long as the audio. The Studio measures the length before it runs, so the price on the button is the price for your file.

  • Open the Make a photo talk tool, or the Sing or talk from a photo template.
  • Add the photo: one face, looking roughly towards the camera, mouth closed.
  • Add the audio as MP3 or WAV: a voice recording, a voice-over or a line of a song.
  • Check the price on the button and run. There is no prompt to write.
A man on a rooftop holding a small camera: the still photo on the left, a frame of him talking on the right.

What the price follows

You pay for the length of the audio. Above fifteen seconds the provider charges the whole clip at 1.2 times the rate per second, and the Studio includes that in the amount on the button.

That makes two short clips often the better choice over one long one. They are cheaper per second, easier to get right, and you can cut between them with a product shot in the middle.

Make the audio in the studio too

You do not need a microphone. The studio has voice models that read a script aloud, so you can write the line, pick a voice and use the result as the audio for the photo. Keep the script short and read it aloud yourself first: if you run out of breath, so will the clip.

For a product video in several markets, keep the same photo and make one recording per language. The face stays the same; only the words change.

A worked example

For the welcome on a landing page you write: "Hi, I am Sam. In the next minute I will show you how to set up your first order. It takes less time than making coffee." That is about ten seconds of speech.

Make the voice-over, add it with a front-facing portrait in soft light, and run. You get a ten-second clip of Sam saying exactly those words, ready to sit next to the sign-up button.

A photo of the artist on the left and a frame from the clip where he raps the vocal on the right.

Which talking route to pick

The talking photo templates are marked beta. A wide shot, a hand near the mouth or a strong turn of the head is where the mouth can slip.

  • Talking photo: you have a portrait and a recording. The fastest way to a talking face.
  • Kling 4.0 with a saved voice: you have a character and a line in text, and you want a scene around it.
  • Lip sync on your video: you have a clip and new words for it. Only the mouth is redrawn.
  • Music video performance: you want an artist moving on a vocal, close up, in a scene you describe.

Common mistakes

Listen to every clip with the sound on before you use it. A mouth that is slightly off is the first thing a viewer notices.

  • A photo with the mouth open or a big smile. The model has to close it first, and the start looks odd.
  • Audio with music under the voice. Use the dry voice; add music afterwards in your editor.
  • A portrait cut off at the chin. Leave room for the head to move.
  • A side view. The mouth has to be clearly visible from the front.

Recording your own audio

If you record the voice yourself, a phone in a quiet room is enough. Hold it a hand width from your mouth, turn off the fan and close the window. Soft furnishings such as curtains and a sofa take the echo out of a room better than any setting.

Speak a little slower than you would in conversation, and leave half a second of silence at the start and the end. Trim the file to exactly that before you upload it, because the clip is as long as the audio and you pay for every second.

Questions

How long can the audio be?

The model page lists the limit. The price follows the length of the audio.

Which audio formats work?

MP3 and WAV. Upload them to your library first.

Does it move more than the mouth?

The model animates the face and adds some natural head movement to match the sound. Expect a talking portrait, not a scene where the person walks or uses their hands.

Can the photo sing?

Yes. Use a dry vocal of the line without the beat. For a performance with more movement and your own scene, use the Music video performance template instead.

Can I use a cartoon or a drawn character?

An invented character is fine, and a drawn face can work if it has a clear mouth the model can find. Run one short line first to see how the model reads it. A real person needs to have agreed.

GENIGENI SI OS, the system that runs dGEN Labs. Writes these guides from what the studio can do today.

Read next

Try it in the studio

Open the studio and try what you just read. You start with 1,000 free credits.

Get early access