How to make a still photo speak a recorded message
Upload the photo and give Vizard Agent the message. Vizard Agent searches for a voice with the right age and warmth, records the words, drives the face from that recording so the mouth matches, checks the result frame by frame for distortion, and cuts it together with music and captions.
What do you type into Vizard Agent?
Describe the voice in human terms. Vizard Agent searches by that description rather than by technical settings, so "an elderly woman, gentle and kind" narrows the search far more usefully than any parameter would, and it will play candidates back before committing. Write the message out exactly as it should be said.
Prompt
What does Vizard Agent actually do?
The technical part is ordinary now. What decides whether the result is worth keeping is the voice and the face holding steady, and both are checked rather than assumed. The long middle of the job was not the talking head at all, but a small watermark across the subject's chest. In the session behind this article it:
- Opened both photos, analysed the supplied audio and searched for an elderly female voice with the right warmth in the target language.
- Generated the narration, evaluated its quality and tone, then drove each photo from the audio and extracted frames to check the face.
- Generated a music bed and transition sound effects, measured every track in LUFS, and inspected the crossover frames between the two people.
- Located a stray watermark by pixel coordinates and tried feathered repair, precise repair and repair from the original image, checking the frame after every attempt, because a repair that is slightly too aggressive smears the face next to it.
- Built animated captions from word-level timings in a font that renders the language, refined the camera move between the two shots and delivered 13.67 seconds at 720x1280.
When does this not work well?
This is the most sensitive kind of video on this site, and whether it should be made at all sits outside the edit. Use only photos of people you have the right to use, and never make someone appear to say something they did not. For someone who has died, the family decides. Low-resolution or angled photos also distort.
Common questions
What kind of photo works best? A clear, front-facing face at reasonable resolution. Angled or blurred photos distort.
Can I use my own voice recording? Yes. Upload the audio and Vizard Agent drives the face from it rather than generating a voice.
Will it look distorted? Vizard Agent extracts frames and checks the face specifically for stretching before it accepts the shot.