Publishing workflow
Your talking photo rendered. It is not ready to publish yet.
Treat the generated MP4 as a draft. Review it like edited media: verify the message, disclose the synthetic performance, and prepare an accessible final cut.
1. Watch with sound, then without sound
First check lip timing, facial artifacts, clipped audio, and whether the person appears to say exactly what was authorized. Then mute the clip: if the message becomes unclear, captions are necessary for comprehension as well as accessibility.
2. Add a visible AI disclosure
Use plain wording such as “AI-animated from a still photo; voice recording supplied by the creator.” Put it in the video or adjacent caption where a viewer will encounter it before drawing the wrong conclusion. Do not hide disclosure in tags.
3. Caption the actual recording
Transcribe what the uploaded speaker says, not what the script was supposed to say. Correct names and punctuation manually. TalkingPhoto does not currently create captions, so add them in your editor or publishing platform.
4. Edit for the destination
TalkingPhoto returns the model’s MP4 rather than a social-media canvas. Use an editor to set the required aspect ratio, safe margins, cover frame, subtitles, and loudness. Export once after those changes to avoid repeated quality loss.
5. Keep evidence of consent
Retain the source permission, final approved message, and where the clip was published. If consent is withdrawn or the context changes, remove the post and any downstream copies you control.