CreationsFavoritesAssetsProjects
All
Image
Video
Features
Back to Agents

Photo Talking Avatar

Create an AI talking avatar from a photo and a script. Use a portrait as the presenter reference, add an optional product photo, or upload only a product image for an AI-generated adult presenter. No video upload is needed.

Required

Upload one clear portrait or product photo in JPG, PNG or WebP, up to 10 MiB. A portrait guides the presenter’s appearance. With a product-only image, the model generates an adult presenter to introduce it. No video is needed.

Optional

Optional: when the main photo is a portrait, add one product photo for the presenter to showcase. JPG, PNG or WebP, up to 10 MiB. Leave empty for a regular talking video or when the main photo already shows the product.

Required
0 / 200
View my creations
Loading your media…

What is a photo talking avatar?

A photo talking avatar is a generated video in which a presenter speaks a script using a still image as a visual reference. Upload a portrait to guide the presenter’s appearance, add a separate product photo for a product introduction, or use only a product image so the model creates an adult presenter. This Seedance 2.5 play needs no presenter video. A single photo does not provide an original voice sample or guarantee exact identity in every generated frame.

Creation examples

Loading your media…

How to make a talking avatar from a photo

Choose your main reference image, add the spoken script and review the video settings.

  1. Choose a portrait or a product as the main photo

    Upload one JPG, PNG or WebP up to 10 MiB. Use a clear portrait to guide the presenter’s appearance, or a product photo if you want the model to create an adult presenter.

  2. Add a script and an optional product photo

    Enter up to 200 characters in the language you want spoken. If the main photo is a portrait, you may add a separate product image for that presenter to showcase.

  3. Select duration, resolution and framing

    Choose 15, 20 or 30 seconds, 720P or 1080P, and 9:16, 16:9 or adaptive framing. Review the estimate before generating and check the result before sharing.

Three ways to use photo talking avatars

Match the image you already have to the type of presenter-led message you want to create.

  • A portrait for a short spoken message

    Use a portrait as the visual reference for a greeting, a short explanation or a creator update. Keep the script focused on one message.

  • A portrait and product photo for an introduction

    Combine a presenter reference with a separate image of the item being introduced. Describe its appearance and verified uses in the script.

  • A product-only image for an AI presenter

    Start with a product image when you have no presenter photo or video. The template asks for an appropriately dressed adult presenter to introduce the item.

What the photo talking avatar workflow supports

One required image and a script are enough to start; a second product image is optional.

  • No presenter video required

    The main reference can be a still portrait or a product photo. This is useful when you have image assets but no recorded presenter clip.

  • A presenter can be generated from a product image

    When the main image contains a product rather than a person, the prompt asks the model to create one adult presenter. The generated person is not a preserved identity from an uploaded portrait.

  • Optional product reference alongside a portrait

    Add a second image when you want the person in the main photo to showcase a specific product. The prompt requests consistent visible details without inventing features or labels.

How to prepare better photos and scripts

Clear inputs make the intended presenter, product and spoken message easier to distinguish.

  • Keep the main reference unambiguous

    For a portrait, keep one face clear and unobstructed. For a product, show one primary item with visible shape and packaging. Avoid collages with several competing subjects.

  • Write a short script in the spoken language

    For 15 seconds, begin with about 20–35 English words or 40–60 Chinese characters. Use short sentences and put the most useful fact early.

  • Inspect generated details before use

    Review pronunciation, lip movement, facial details, hands and product labels. Confirm that spoken product claims match your actual product information.

Frequently asked questions

Do I need to upload a video?

No. This play uses one required main photo and a script. A separate product photo is optional. Use the video-based AI Talking Avatar play if you want to provide a presenter clip instead.

Can I use a portrait and a product photo together?

Yes. Put the portrait in Main photo and the product image in Product photo. The template asks the person referenced by the portrait to showcase and introduce that product.

Can I generate a presenter from only a product image?

Yes. Upload the product image as the required Main photo and leave the optional Product photo empty. The model is asked to create one appropriately dressed adult presenter to introduce the item.

Which image formats and sizes are supported?

Each image must be a JPG, PNG or WebP up to 10 MiB. Use a clear single-person portrait or a photo with one primary product; avoid small, blurred or heavily obstructed subjects.

Which durations and resolutions can I choose?

The draft offers 15, 20 or 30 seconds, 720P or 1080P, and 9:16, 16:9 or adaptive framing. Review the available settings and current estimate in the generation form.

How should I write Chinese or English scripts?

Enter up to 200 characters in the language you want spoken. For 15 seconds, start with about 20–35 English words or 40–60 Chinese characters. Use short sentences and verified product facts.

Will a photo reproduce the person’s exact voice and appearance?

A still photo provides no original voice sample. It guides appearance, while the model generates speech and movement. Exact identity, voice and lip synchronization are not guaranteed, so review the result.

Will the product and its labels stay exactly the same?

The prompt asks the model to preserve visible shape, colors, packaging and existing branding, but generated details can vary. Check labels, logos, hands and spoken claims before using the video.

Drag the cat to move it. Click to contact support.