1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
const axios = require('axios');
const api_key = "YOUR API-KEY";
const url = "https://api.segmind.com/v1/veed-2-lipsync";
const data = {
"video_url": "https://segmind-resources.s3.amazonaws.com/input/veed-2-lipsync-input-video-20260717.mp4",
"audio_url": "https://segmind-resources.s3.amazonaws.com/input/veed-2-lipsync-input-audio-20260717.mp3"
};
(async function() {
try {
const response = await axios.post(url, data, { headers: { 'x-api-key': api_key } });
console.log(response.data);
} catch (error) {
console.error('Error:', error.response.data);
}
})();Attributes
Public URL of the source talking-head video. Use a clear, front-facing, well-lit face.
Public URL of the driving speech track. Use clean voiceover or TTS for tighter sync.
To keep track of your credit usage, you can inspect the response headers of each API call. The x-remaining-credits property will indicate the number of remaining credits in your account. Ensure you monitor this value to avoid any disruptions in your API usage.
VEED Lipsync v2 — Video-to-Video Lip Sync & Dubbing
What is VEED Lipsync v2?
VEED Lipsync v2 is a video-to-video lip sync model that re-syncs a talking-head video to a new speech audio track. You provide a source video and an audio file, and the model detects the on-screen face and re-renders the mouth so the speaker appears to naturally say the new audio. Unlike first-generation lip sync, VEED Lipsync v2 transfers emotion, speaking style, and timing directly from the audio, so the output reads as the speaker's own performance rather than a generic AI overlay. It is zero-shot: there is no per-subject training, reference footage, or fine-tuning, so it works on any new face the first time — ideal for automated dubbing and localization pipelines.
Key Features
- Emotion and speaking-style transfer straight from the driving audio.
- Zero-shot lip sync that works on any face with no training step.
- Video-to-video on real footage, keeping your original take intact.
- Language-agnostic audio input for multilingual dubbing and localization.
- Simple two-input API:
video_urlandaudio_url, returning a finished MP4.
Best Use Cases
- Dubbing and localization: pair the source video with translated voiceovers to ship market-ready versions without reshooting.
- Ad iteration and A/B testing: swap in a new hook on the same take.
- Line fixes: replace a flubbed sentence without a re-record.
- AI presenters and avatars: drive existing footage with fresh scripts.
In testing, a front-facing talking-head clip re-synced cleanly to a fresh TTS track with natural mouth movement and no visible artifacts.
Prompt Tips and Output Quality
VEED Lipsync v2 is driven by media, not text prompts. For the tightest sync, use a clear, forward-facing, well-lit clip with the mouth visible throughout, and clean speech audio. Keep the audio no longer than the source video, since the output duration follows the audio. Best results come from a single active speaker; non-human subjects are unsupported.
FAQs
How is VEED Lipsync v2 different from the original lip sync? It adds emotion and style transfer, higher fidelity, and support for longer, higher-resolution clips.
Do I need to train a model per speaker? No. It is zero-shot and works on any new face immediately.
Can I use it for localization? Yes. Provide translated audio in any language to produce localized versions.
What footage works best? A single, forward-facing human speaker with a clear, well-lit, unobstructed face.
What do I send the API? Two inputs — a source video URL and a speech audio URL.
Popular Models
