Validation
The upload path checks ownership, type, size, and image readability before proceeding. A clear, visible face is still required for the talking-head model to work well.
Essential storage keeps your sign-in and preferences working. With your permission, our analytics providers help us measure page and conversion performance; advertising storage stays off.
Loading…
Portrait + speech audio → MP4
TalkingPhoto is a focused online workflow for animating one visible face from a speech recording. It does not pretend to offer stock avatars, voice cloning, or text-to-speech when those services are not configured.
JPG, PNG, or WebP up to 10MB.
MP3, WAV, or M4A up to 10MB.
A downloadable MP4 after successful asynchronous processing.
The upload path checks ownership, type, size, and image readability before proceeding. A clear, visible face is still required for the talking-head model to work well.
A credit is reserved and an asynchronous Inngest task calls the pinned SadTalker model version on Replicate.
The completed provider file is copied to object storage. Owner-checked API routes expose status and download access.
Create an account, upload one clear portrait, upload your speech recording, and submit a generation job. TalkingPhoto validates the files, sends the inputs to an AI talking-head model, and stores the resulting MP4 when generation succeeds.
Not in the current production configuration. TalkingPhoto currently requires an MP3, WAV, or M4A recording so the displayed workflow matches the services that are actually configured.
No. The model needs a clearly visible face, and output quality varies. Clear front-facing portraits with unobstructed eyes and mouth are more reliable than small, blurred, heavily shadowed, or side-profile faces. The current upload step checks readability rather than guaranteeing automatic face detection.
No. AI talking-head output can show lip-sync, teeth, eye, skin, or head-motion artifacts. Review the generated file before publishing it.
The first private preview lets a new account test the real workflow without inventing a testimonial or demo result.
Contact us