Discussions

Hi team,

I'm using the HeyGen Video Creation API to generate a video where a 3D avatar transitions between idle and talking states using silence segments and voice input. Here’s a simplified version of my video_inputs:

"video_inputs": [
{
"character": {
"type": "avatar",
# "scale": 1,
"avatar_style": "normal",
"avatar_id": avatar_id
},
"voice": {
"type": "silence",
"duration": "2.0"
}
},
{
"character": {
"type": "avatar",
# "scale": 1,
"avatar_style": "normal",
"avatar_id": avatar_id
},
"voice": voice
},
{
"character": {
"type": "avatar",
# "scale": 1,
"avatar_style": "normal",
"avatar_id": avatar_id
},
"voice": {
"type": "silence",
"duration": "5.0"
}
}
]

The issue is that transitions between idle → talking or talking → idle are not smooth — it’s clearly visible that the video is switching or "cutting." This makes the avatar movement feel unnatural.

Question:
Is there any recommended way to make these transitions smoother? For example:

Should I use a specific avatar_style or parameter?

Is there any fade/transition support between segments?

Would adding overlapping frames or modifying silence durations help?

Any best practices or tweaks to make the transitions more natural would be highly appreciated.

Thanks in advance!