How to Make an AI Avatar Video from One Photo and a Script
An AI avatar video makes the person in a single image speak, with the mouth, face and head following a voice track. You can have a character or a stand-in of yourself deliver announcements, explainers or greetings without filming anything. This guide uses BunCho's AI avatar tool to cover choosing the image and voice, picking a model, the steps, fixes for common problems and the rights you need.
What an AI avatar tool does
You provide an image of a person and a voice — an audio file, or a script read aloud by a built-in voice. The tool makes a video of that person speaking, with the mouth, face and head moving to the voice. The length matches the voice: 2 to 30 seconds in BunCho.
Typical uses are character announcements and greetings, a narrator for explainers without showing your face, and short clips where a story character says their lines in the original art. If you make manga panels or character art in the studio, you can make those speak directly.
- Input: one image of a person and a script (up to 300 characters) or audio (up to 30 s)
- Output: an MP4 of that person speaking (2–30 s)
- Models: Kling Avatar 2.0 Pro / Standard, OmniHuman 1.5, HeyGen Avatar IV, P-Video Avatar
What to prepare (image and voice)
- Image: one person with a clearly visible face. Front-facing images where the face is large give the cleanest mouth movement
- Format and size: PNG, JPEG or WebP; short side at least 300 px; aspect ratio between 1:2.5 and 2.5:1
- Voice from a script: up to 300 characters, read by one of four voices (bright female, calm male, female narrator, male narrator)
- Voice from a file: up to 30 seconds. Speech without music or background noise works best
- Extra direction (optional): describe the expression or framing, e.g. "smiling, looking at the camera"
How to make an AI avatar video
-
1Open the video tools
In the studio, open Video tools and choose AI avatar.
-
2Add the image
Upload one image under Person image.
-
3Set the voice
Type the lines and choose a voice, or switch to audio file and upload your recording.
-
4Pick a model and estimate
Choose a model and press Estimate. The credits for each model are listed so you can compare.
-
5Confirm rights and generate
Tick the box confirming you have the rights to the image and audio, then press the generate button. It can take a few minutes; the job continues if you close the page.
-
6Download and use
Download the finished video and add captions or music in your video editor.
See also: AI avatar tool overview
Models you can choose, and approximate credits for 5 seconds
Kling Avatar 2.0 comes in Pro and Standard; Standard uses about half the credits of Pro. OmniHuman 1.5 also moves the hands and body but uses more credits. HeyGen Avatar IV outputs 1080p, and P-Video Avatar outputs 720p with the fewest credits.
| Model | Length | Approx. credits | Difference |
|---|---|---|---|
| Kling Avatar 2.0 Pro | 2–30 s | about 438 for 5 s | High quality, takes a few minutes |
| Kling Avatar 2.0 Standard | 2–30 s | about 221 for 5 s | About half the credits of Pro, slightly lower quality |
| OmniHuman 1.5 | 2–30 s | about 667 for 5 s | Also moves hands and body, more credits |
| HeyGen Avatar IV | 2–30 s | about 194 for 5 s | Fewer credits, 1080p |
| P-Video Avatar | 2–30 s | about 126 for 5 s | Fewest credits, fast, 720p |
Fixing common problems
- Mouth movement is small or off: small faces or side views do not work well. Crop to a large, front-facing face
- The voice sounds muffled or cuts out: remove music and noise so the file contains only speech
- Motion looks unnatural for the art style: illustration and anime styles vary. Test with a low-credit model before the final version
- Stopped with an image size error: the short side must be 300 px or more and the aspect ratio between 1:2.5 and 2.5:1
- You need more than 30 seconds: make it in 30-second parts and join them in your editor
Rights and AI disclosure
Only use images, audio and video that you own or have permission to use. Real people's faces and voices may only be used with that person's permission. Before generating, you tick a box confirming you have the rights.
You may not make videos that show someone saying something they did not say in a misleading way, or that impersonate someone. When you publish, label the video as AI-generated or AI-edited according to the platform's rules (for example, YouTube's altered or synthetic content setting).
Model names in this guide and in the product are listed because you can choose them in BunCho. Listing them does not imply any partnership with or endorsement by their makers.
Plans and credits
The video tools are available on paid plans (Watering, Seed, Feast, Full Course, Team & Business) with an active subscription. They are not available on the free plan or during a paid plan's free trial.
They use your studio's monthly AI credits, the same balance as novels, manga and video. You see an estimate before generating and nothing is made until you agree. No credits are used if a job fails or if you cancel it before it starts.
See also: Plans and pricing / All five video tools
FAQ
How long can an AI avatar video be?
It follows the voice: 2 to 30 seconds. Scripts up to 300 characters, audio files up to 30 seconds.
Is it free?
The video tools are available on paid plans with an active subscription. They are not available on the free plan or during the free trial.
Can it speak with my own voice?
Yes. Upload your own recording of up to 30 seconds as an audio file.
Can I use a photo of a real person?
Only with that person's permission. Photos used without permission, or uses that impersonate someone, are not allowed.
Pages we checked for this guide
- BunCho video tools: AI avatar bun-cho.work · primary source
- BunCho plans and pricing bun-cho.work · primary source
Features, prices and screens are as of October 30, 2026 and may change. Images are actual BunCho screens (some show the Japanese interface) and samples made in the Studio. Japanese original