AI Talking Avatar API from a Reference Video

Send a video of a person and a speech track to POST /api/v1/videos/generations with model: ai-talking-avatar-video. The whole frame is regenerated with your video as the reference, so head movement and facial expressions follow the speech, not just the lips. Output is 540p or 720p, from 3 to 60 seconds.
Get API Key
Softly lit presenter set with a camera and teleprompter facing an empty stool

A Full Performance, Not Just Lips

Expressions follow the speech

Because the whole frame is regenerated, head motion and facial expression react to the audio. Add an optional prompt such as a man speaking calmly to the camera to set the tone.
Row of 3D rendered head sculptures with different expressive tilts

Choose the length strategy

video_mode: normal (default) uses the shorter of video and audio; loop follows the audio and plays a shorter video forward and backward to cover it.
Hourglass beside an audio waveform printed on paper

Cut it short with duration

Set duration (3 to 60 whole seconds) to use only the first N seconds and pay for N seconds. With duration set, the source files may be longer than 60 seconds.
Pair of scissors next to a strip of film on a cutting mat
Both files are measured before tokens are deducted; that measurement enforces the 3 to 60 second limit.

AI Talking Avatar (Video) Request Fields and Limits

model

Always ai-talking-avatar-video.

video_url (required)

Video of the person, up to 300 MB. One person, face clearly visible and roughly front-facing for most of the clip. The output keeps this framing.

audio_url (required)

Speech audio, up to 50 MB. Clear speech with little background noise works best.

resolution

540p (default) or 720p.

video_mode and duration

normal (default) or loop; optional duration is an integer from 3 to 60 and cannot exceed the output length set by the mode.

prompt

Optional description of the performance. Defaults to person talking.
Billed per second of output; 720p costs 1.5 times as much as 540p.

AI Talking Avatar (Video) API Pricing

540p: 1 token ($0.02) per second

A 30-second clip costs 30 tokens ($0.60).

720p: 1.5 tokens ($0.03) per second

A 30-second clip costs 45 tokens ($0.90).

60-second maximum

The longest output costs 60 tokens ($1.20) at 540p or 90 tokens ($1.80) at 720p.

Billed on measured length

Without duration, the bill uses the measured output length for your video_mode; with it, exactly duration seconds. Failed jobs are refunded automatically.

Integrate the AI Talking Avatar (Video) API

1

1. Create an API key

Generate a key at /keys and send it as Authorization: Bearer <key> from your server.
2

2. Record or pick the reference

Use a clip with one front-facing person and a clean speech track; choose loop if the audio is longer than the video.
3

3. Submit the job

POST {"model": "ai-talking-avatar-video", "video_url": "https://…", "audio_url": "https://…", "resolution": "720p"} to /api/v1/videos/generations.
4

4. Collect the video

Store img_uuid, poll GET /api/v1/jobs/detail or wait for your callback, and download the result before the link expires.
Create an API Key
Natural-looking spokesperson clips from short reference recordings.

What Developers Build with AI Talking Avatar (Video)

Spokesperson and explainer videos

Record a presenter once and generate new announcements from text-to-speech or voice recordings.

Course and training narration

Turn a short loopable clip of an instructor into lesson videos that follow the narration.

Social and UGC avatars

Let users animate a selfie video with a voice line, with expressions that match the delivery.

Multilingual versions

Pair translated audio with the same reference video to produce localized talking clips.
Inputs, length, pricing and related models.

AI Talking Avatar (Video) API FAQ

What is the AI Talking Avatar (Video) API endpoint?

POST https://api.goenhance.ai/api/v1/videos/generations with model: ai-talking-avatar-video.

How is it different from lipsync-video?

This model regenerates the whole frame so the head and face move with the speech. lipsync-video redraws only the mouth and keeps every other pixel.

Can I start from a photo?

Not with this model. For a single photo, use ai-talking-avatar.

How long can the output be?

3 to 60 seconds. With duration set, the source files themselves may be longer than 60 seconds.

How is it priced?

1 token ($0.02) per output second at 540p and 1.5 tokens ($0.03) per second at 720p.

Can I swap the URLs after submitting?

No effect. Both files are copied to GoEnhance storage and measured before any tokens are deducted.

Start Building with the AI Talking Avatar (Video) API

AI Talking Avatar Video API: Pricing & Integration | GoEnhance