> ## Documentation Index
> Fetch the complete documentation index at: https://docs.kymaapi.com/llms.txt
> Use this file to discover all available pages before exploring further.

# Kling AI Avatar v2 Pro

> Kling AI Avatar v2 Pro is Kuaishou’s audio-driven avatar model. Give it one portrait and one speech or song track, and it returns a clip of that face lip-syncing to your audio.

## Overview

Kling AI Avatar v2 Pro, from Kuaishou’s Kling Avatar 2.0 line (late 2025), animates a single still image from an audio track. The mouth follows the speech or singing, the head and shoulders move naturally, and the audio you send plays through unchanged. There is no duration setting: the clip runs as long as the audio.

In our own tests it kept a synthetic singer’s identity, lip-synced a Vietnamese vocal with the mouth closing on breaths, and kept the original vocal in the soundtrack. A 13 second clip took about six minutes. Kyma accepts uncompressed PCM WAV audio of at least 5 seconds (WAV PCM only for now).

Billing is per second of the finished clip. You declare your audio’s length as duration\_s; Kyma checks the audio before generating, cuts anything longer, and charges the clip’s real length.

## Specs

| Field | Value |
| - | - |
| Model ID | `kling-avatar-v2-pro` |
| Best for | Talking-head clips, lip-synced singing, a narrator or presenter from a photo and a voice track |
| Input modalities | Text, Image, Audio |
| Output modalities | Video, Audio |
| Speed | Slow |
| Cost band | Premium |
| Release stage | Stable |

## Pricing

| | Price |
| - | :-: |
| Per second of video | \$0.15525 |

Billed per second of generated video. Every request reports its exact cost in `usage.cost`; the full media price list is on [Pricing](/pricing).

## Use this when

* **Talking-Head Clips** — Turns a presenter photo and a voice recording into a speaking clip for explainers, lessons and announcements.
* **Lip-Synced Singing** — Makes a character sing your own song track, keeping the real vocal in the soundtrack.
* **Any-Language Voiceover** — Lip-syncs to the audio itself rather than to a transcript, so the language of the recording does not need to be supported separately.
* **Character Series** — Reuses one still of a character across many clips, each driven by a different line of audio.

## Not ideal for

Do not use this model for text-to-video scenes, camera moves, or clips longer than 60 seconds; it animates one face from one image and needs an audio track.

## Example

```python theme={null}
from openai import OpenAI

client = OpenAI(base_url="https://kymaapi.com/v1", api_key="kyma-...")

# Video is POST /v1/videos/generations — it returns a job to poll.
response = client.post(
    "/videos/generations",
    body={"model": "kling-avatar-v2-pro", "prompt": "..."},
)
```

## Agent query example

Ask the API which models fit, instead of hardcoding an id:

```bash theme={null}
curl "https://kymaapi.com/v1/models?recommended_for=generate-video&quality_tier=frontier-open"
```

## FAQ

**What do I send?**

image\_url (the face), audio\_url (the speech or song as a PCM WAV file, at least 5 seconds) and duration\_s, your audio’s length in seconds rounded up. A prompt is required and steers style and motion.

**How long is the clip and what does it cost?**

As long as your audio. You are charged per second of the finished clip; duration\_s caps the charge, and audio longer than duration\_s is cut before generation.

**Does it keep my audio?**

Yes. The clip plays the audio you sent; the model only animates the face to match it.


This documentation is built and hosted on [Mintlify](https://mintlify.com), a developer documentation platform.