HeyGen product update: Avatar IV, memory, and Veo 3.1 integration!
HeyGen recently hosted a product update webinar, highlighting significant advancements in its AI video creation platform, focusing on enhancing realism, control, and creative possibilities for users' digital twins.
Here's a summary of the exciting new features and updates:
Upgraded Avatar IV model
The core Avatar IV model has received massive improvements in both its Fast Mode and Quality Mode.
- Fast mode improvements: Now offers improved realism, fidelity, and clarity, alongside exceptional stability. Users can confidently create long videos (e.g., 10 minutes) in a single shot without needing constant re-dos for glitches. It's also 50% faster.
- Quality mode enhancement (timing): The quality mode offers more expressiveness, now has an improved sense of timing. This is crucial for custom motion prompting. Previously, a gesture like a "wave" might repeat throughout a long scene; now, the model understands the timing, allowing for specific sequences of gestures that occur only when prompted (e.g., a wave at the beginning, followed by natural talking).
New expressiveness control
Users now have more granular control over their avatar's movements with a new expressiveness control setting.
- Adjustment: You can lower or increase the expressiveness level.
- Default setting: The default is recommended for most use cases, as it's been refined for great results.
- Lower expressiveness: Recommended if the source photo has an overexaggerated expression (like a huge smile) or for maximum stability in long scenes. Lower expressiveness is more stable and is currently rated highest by users for overall usability.
- Higher expressiveness: Useful for serious input photos, or for use cases requiring a super animated, dynamic, or engaging style, such as UGC product ads. Note that higher expressiveness can increase the possibility of artifacts.
- Availability: Supported for photo-only avatars in both fast and quality mode for now, with plans to support digital twins later.
Introducing avatar memory
Avatar Memory is a major upgrade to the Avatar Unlimited mode, designed to make your avatar better and more reusable over time.
- How it works: Every time you generate footage with Avatar IV that meets the quality requirements, that motion footage is saved to your avatar's memory.
- Unlimited reuse: You can then reuse those exact motion clips (like a favorite introduction wave or a natural talking style) in the unlimited mode for future videos without consuming monthly minutes.
- Unlimited use: Create unlimited videos by reusing previously generated motion clips.
- Near-instant previewing: Clips generated from memory preview almost instantly in the studio, speeding up the iteration process.
- Consistency and personality: Over time, you build a rich library of consistent, unique gestures and expressions, helping to craft a cohesive and memorable personality for your virtual character.
- Trade-off: Only the lip sync adapts to the new script; the rest of the body movement loops, so it's essential to preview the clip in the studio to ensure the motion is in sync with the audio.
Veo 3.1 integration
HeyGen now integrates with Google's state-of-the-art video generation model, Veo 3.1, bringing new dynamic capabilities while maintaining your avatar's identity.
- Identity consistency: The HeyGen integration provides voice and appearance consistency to Veo 3.1 generations. You can use your existing avatar and voice, even as you generate new, dynamic scenes.
- Multi-character dialogue: Your avatar can now have a dialogue with another character (prompted or referenced).
- Dynamic scenes: Easily create scroll-stopping hooks for social media (e.g., avatar smashing through a wall) to grab attention.
- Consistent voice and look: You can prompt for your avatar to be in different outfits, backgrounds, or poses while keeping the same voice and appearance.
- Longer generations: Generate multi-scene generations up to 60 seconds long with seamless transitions.
- Reference to video: A specific shortcut allows you to upload multiple reference images (people, products, environments) to introduce elements or characters mid-scene or include specific objects that are not already a part of the avatar.
HeyGen Boot Camp announcement
HeyGen is hosting a Boot Camp, a 2-day virtual creative masterclass scheduled for December 3rd and 4th. The camp is designed for anyone looking to seriously scale their video strategy across various use cases like courses, social media, and executive communications.
