We’re excited to launch the Video Agent endpoint (/v1/video_agent/generate), a powerful "one-shot" tool that creates high-quality avatar videos from simple, natural language prompts.

Instead of manually configuring every scene, script, and asset, you can now provide a description of what you want. The Video Agent handles the heavy lifting—from scriptwriting to visual assembly—allowing you to scale video production with minimal overhead.

What’s New:

• Prompt-to-Video: Generate a complete video from a single text description.
• Reduced Complexity: No need to orchestrate complex JSON structures for scenes or assets.
• Rapid Automation: Ideal for personalized messaging, social content, and internal comms at scale.

📘

Ready to test it out?

Head over to the Video Agent API Reference to try a request directly from your browser.

We now support voice configuration for ElevenLabs V3 Voice Model in Create Avatar Video API.

This is in addition to the existing supported models: eleven_monolingual_v1, eleven_multilingual_v1, eleven_multilingual_v2, eleven_turbo_v2, and eleven_turbo_v2_5 for video generation.

To use ElevenLabs V3 model, please set

voice.elevanlabs_settings.model (string, optional): to be eleven_v3

Voice.elevanlabs_settings.stability (float, optional): default value is 1.0, allow value to one of 0, 0.5, 1.0

We’ve introduced a new Video Translation model, quality, designed to deliver significantly more natural lip-sync and context-aware translations. This new mode is optional and available via the /v2/video_translate endpoint using the new mode parameter.

Available Modes

Fast Mode (default)

mode: fast

Cost: 3 credits per minute

Optimized for speed and standard translations, especially effective for videos with limited facial movement.

Quality Mode (new)

mode: quality

Cost: 6 credits per minute

Produces highly natural, context-aware lip-sync for premium translation results.

You may now specify the desired mode when calling the /v2/video_translate endpoint. If no mode is provided, it will default to fast.

Reference: https://docs.heygen.com/reference/video-translate#/

We’ve added a new endpoint to retrieve dynamic and custom variables defined in your AI Studio templates.

Endpoint: v3/template/<template_id> This endpoint allows developers to programmatically access all variable definitions — including avatar, text and elements and media (e.g. image, video, etc.) — that can be dynamically populated during video generation. It enables full integration of AI-powered avatar video creation workflows using your own template schema.

Reference: https://docs.heygen.com/reference/get-template-v3#/

We've added the new endpoint to Create Avatar IV

Endpoint: v2/video/av4/generate - To enable programmatic generation of AI-powered avatar videos using advanced photorealistic avatar technology. This endpoint allows you to create videos with custom avatars speaking your provided text or audio.

Reference: https://docs.heygen.com/reference/create-avatar-iv-video

We’ve expanded the Asset API with new endpoints for easier asset management:

• Endpoint: /v1/asset/list — List assets with pagination and filters for file_type and folder_id.
  Reference: https://docs.heygen.com/reference/list-assets
• Endpoint: /v1/asset/<asset_id>/delete — Delete a single asset by its ID.
  Reference: https://docs.heygen.com/reference/delete-asset

Endpoint: /v1/streaming.new
Reference: https://docs.heygen.com/reference/new-session

Added new STT provider: AssemblyAI. It is now the default provider for English language transcription.

Endpoint: /v2/video/generate
Reference: https://docs.heygen.com/reference/create-an-avatar-video-v2#subtitles-object

You can now customize subtitles in avatar videos using the new text object on the /v2/video/generate endpoint. This gives you more flexible, dynamic subtitle options—including preset styles, custom font sizes, position control, alignment, etc.

Endpoint: /v1/streaming.new
Reference: https://docs.heygen.com/reference/new-session

A new parameter activity_idle_timeout has been added to the streaming.new endpoint.
It specifies the maximum idle time (in seconds) after the last user interaction before the session is marked as inactive.Default: 120 seconds, Range: 30 to 3600 seconds. This gives you greater control over session lifecycle management and resource usage.

Endpoint: /v1/streaming.keep_alive
Reference: https://docs.heygen.com/reference/keep-alive

A new /v1/streaming.keep_alive endpoint has been introduced to reset the idle timeout counter for an active streaming session. Use this endpoint to keep the session alive and prevent it from timing out due to inactivity.