Skip to main content
Avatar video turns a single face photo plus a script — or your own audio track — into a talking-head clip. Pick an avatar model with model on generate_video; the shared arguments are on Generate Video with AI.

Avatar models

conn.get_generation_models() lists the models your account can use. More models will be added to this category; the arguments below are per model.

HeyGen (model="heygen")

Access: Pro plans. prompt is the exact script the avatar speaks, word for word. image_url is a public photo with one clear face.
A plain dict or object with the same keys works as config too. Keys are snake_case in Python and the REST API (audio_id, talking_style); the Node SDK takes camelCase (audioId, talkingStyle) and sends them as snake_case.

config

Unknown keys and values outside these lists are rejected before anything is generated, with the allowed values in the message.

Price and time

  • $0.50 per second of output video, billed on the actual length, rounded down to whole seconds. The same at every resolution.
  • Your balance is checked against the script’s estimated length (or the audio’s length) before generating.
  • A short clip takes about 2 to 5 minutes. For long scripts, pass callback_url rather than waiting on the call.

Voices

HeyGenConfig.voices in the SDKs holds a starter set. Every voice:
Some names end with a space or contain an emoji. Copy them exactly.

Use VideoDB in your agent

VideoDB lets an AI agent watch, search and clip video.Free to start: $20 of credits, no credit card. Steps for every client