Talking Photo
模型
PORTRAIT PHOTO
JPG/PNG/WEBP up to 5MB, min 300px
What should it say?
画面比例
时长
最近生成
Related workflows
Continue to a broader photo animation workflow, an old-photo memory clip, or a prompt-based video when your input or goal changes.
Animate a real photo when broader motion matters more than speech.
Use restrained movement for scanned family and historical portraits.
Add motion to artwork, products, and scene images.
Start from a written scene when you do not have a source photo.
Talking photo AI
Animate Photo AI's Talking Photo tool turns one clear portrait and a typed line into a short lip-synced video. Enter up to 200 characters, choose a compatible model with audio, and review the mouth, eyes, and identity before export. Audio-file upload and voice selection are not yet connected in this build.
Real product example
This four-second clip is the portrait-motion example currently published for this workflow. It demonstrates facial stability and subtle movement; it is not presented as evidence of generated speech. The workbench sends a typed line to a compatible audio-capable model for speech-style lip sync.
Script guide
Enter the words the portrait should say, not camera directions. This workflow limits typed scripts to 200 characters. One short sentence is easier to time and review than several unrelated ideas in the same clip.
1
Use a greeting, announcement, or short message. Split longer narration into separate clips.
2
Commas and full stops make the intended rhythm clearer. Review names and numbers before generation.
3
Credits vary with the selected model, duration, resolution, and audio setting. The workbench shows the estimate before submission.
Core capabilities
The page is designed for a controlled first draft from a still portrait. Each capability maps to a concrete input or review decision, so you can tell whether a weak result comes from the source photo, the requested motion, or the output settings.
The talking photo workflow starts with a clear portrait and a restrained prompt for small mouth movement, blinking, or breathing. A close, front-facing input gives the model more stable facial detail than a distant group shot.
Choose a connected video model, aspect ratio, duration, resolution, and sound setting in the workbench. The credit estimate updates before generation so you can test a short draft before committing to a larger output.
The result appears in the preview panel with its status and download action. Compare the generated mouth, eyes, hairline, clothing, and background with the source image before you publish or share the clip.
Enter a short line and use a compatible audio-capable model to create speech-style mouth movement. Audio-file upload and independent voice selection are not yet connected in this build.
Four-step workflow
A reliable talking photo starts with a stable portrait, one motion goal, a short test, and a careful comparison. This order limits unnecessary identity changes before you spend more credits.
Use a JPG, PNG, or WebP image up to 5MB with at least 300 pixels. A front-facing face, visible mouth, even lighting, and an uncluttered background make the first draft easier to judge.
Type up to 200 characters as the words the portrait should say. Keep one idea per clip so the lip sync and facial expression remain easy to review.
Select a model with audio support, then confirm duration, resolution, sound, and the credit estimate before generating. The exact controls depend on the selected model.
Check the mouth shape, eyes, identity, hairline, hands, clothing, and background. Keep the source visible while reviewing so a convincing motion effect does not hide an unwanted change.
Input guide
Match the motion request to the information visible in the source. A talking photo AI workflow cannot reliably infer details hidden by blur, hair, hands, or a strong profile angle.
| Source photo | Start with | Avoid first | Review |
|---|---|---|---|
| Clear front-facing headshot | Small mouth movement and one blink | Large expression or head turn | Mouth, eyes, hairline |
| Avatar or illustrated face | Low-motion presenter test | Redesigning the character | Eye shape, silhouette, colors |
| Family or memory portrait | Respectful facial motion | Claiming spoken words | Identity, clothing, context |
| Side profile or covered mouth | Crop or replace the source | Speech-like mouth motion | Mouth edges and drift |
Published specifications
Use these published limits to prepare the first talking photo draft. Model-specific duration, resolution, and credit values remain visible in the workbench before generation.
Use cases
Use the page when the job starts with one portrait and a short line of dialogue. Choose another workflow for long narration, uploaded voice recordings, or full-scene animation.
Start with a clean headshot and enter a concise welcome, lesson introduction, or product message. Review pronunciation, lip sync, and identity before using the result as a presenter clip.
Upload a readable illustrated or generated face, add one character line, and keep the design unchanged. Check the eyes, mouth edges, colors, and silhouette before publishing.
Use a respectful family portrait with low motion for a private keepsake. Keep the original still available and avoid presenting generated movement as a recording of what the person actually said.
Boundaries and quality control
The current talking photo page is intentionally honest about what a single still image can support.
This workflow does not restore blur, rebuild missing facial detail, or colorize an old photo before animation.
Keep the typed line short and review lip sync frame by frame. Audio-file upload and independent voice selection are not yet connected.
Side profiles, covered mouths, tiny faces, and strong occlusion can cause mouth or identity drift.
Review every generated frame for changes to the face, hands, clothing, and background before publishing.
Only use photos and likenesses you have permission to process. Do not present generated motion as a real recording or statement.
FAQ
Upload one portrait and enter a short line. A compatible audio-capable model turns the line into speech-style mouth movement while preserving the face, framing, and background.
The current workflow uses a typed line and a compatible audio-capable video model for speech-style lip sync. Uploading your own audio recording is not yet connected in this build.
Use one sharp, clearly visible, mostly front-facing face with the mouth uncovered. Portraits, selfies, old photos, drawings, and pets can work; tiny faces and strong side profiles are less reliable.
Upload a clear portrait, type up to 200 characters, choose an audio-capable model, confirm the credit estimate, and generate. Review the mouth, eyes, identity, and background before exporting.
A talking avatar and talking photo both synchronize a face to speech. This page supports a typed script with a compatible audio-capable model; audio-file upload and independent voice selection are not yet connected.
Related next steps
Need broader motion? Continue with Photo to Video. Working with a scanned family portrait? Start with Animate Old Photos and clean the source first.
Upload a clear portrait above, keep the first prompt small, and compare the result with the original before you decide whether the visual motion is ready to use.
Last updated: