What's New in Voice Settings
Important Changes in Studio
No More "Auto" Engine: Studio now explicitly displays the active engine driving your avatar's voice (e.g., ElevenLabs V3, Orca, Panda, Fish).
Consistent Voice Across Your Project – Changing the voice engine updates every scene that uses the same avatar, ensuring a consistent voice throughout your video.
Dedicated "Improve Delivery" Tab: The Improve Delivery tab provides a dedicated space to enhance speech quality and alerts you before switching to a different voice engine.
Quick Revert Option: Easily restore your previous voice engine and settings if you're not satisfied with the new voice output.
How Voice Engines Work
In Studio, a Voice Engine is the underlying AI technology that generates your avatar’s spoken dialogue from your script.
To ensure consistent voice quality throughout your video, HeyGen Studio applies the selected AI voice engine across all scenes that use the same avatar. The Voice panel always displays the active voice engine (such as ElevenLabs V3, Orca, Panda, or Fish), replacing the previous "Auto" label so you always know which engine is generating your speech.
How to Select a Voice Engine
Open your project in Studio.
Click on the Voice panel in the left sidebar or top control bar.
Click the Engine dropdown menu.
Select your preferred engine (e.g., ElevenLabs V3, Fish, Panda, Orca).
Default Settings vs. Video Settings:
Note: Video Settings only apply to the current project. Changing the AI voice engine updates all scenes that use the same avatar within that video. To make this voice setup the default for future projects, click Update Default Settings.
Types of Voice Engines
Each AI voice engine is optimized for different speaking styles, emotional expression, and audio quality. Choose the voice engine that best matches your video's content and the delivery you want to achieve.
Voice Engine | Role | Key Characteristics | Ideal Use Case |
ElevenLabs & ElevenLabs V3 | Studio Realism & Dubbing | Delivers ultra-realistic, natural voiceovers across 70+ languages. By default, uses the Multilingual V3 model for subtle human expressiveness and audio tags. | • High-end promotional videos • Executive presentations • Global multilingual dubbing |
Panda | Expressive Control & Direction
| HeyGen’s core engine built for Voice Director and Voice Mirroring. Adapts vocal delivery dynamically based on natural language prompts (e.g., "emphasize this" or "add excitement"). | • E-learning & tutorials • Marketing & ad content • Expressive avatar performances |
Starfish | Asian Language Optimization | Purpose-built for Asian languages (Chinese, Japanese, Korean, etc.), ensuring natural pacing, pitch accents, and localized pronunciation. | • Asian-market video localization • Region-specific avatar narration |
Cartesia | Low-Latency Streaming | Optimized for fast streaming and sub-second latency, making voice generation quick and responsive. | • Interactive video avatars • Fast content iteration |
Dolphin | Speed & Standard Generation | A lightweight "Turbo" style engine optimized for quick rendering speeds and low credit consumption. | • Draft previewing • High-volume corporate announcements |
Mockingbird | Tone Adaptation & Mimicry | Engineered to match custom vocal styles and adapt dynamically to specialized script tones. | • Character-driven avatar videos • Custom persona voices |
What is "Improve Delivery"?
Improve Delivery is a feature within Voice Settings that helps you enhance your AI voice's speech cadence, emotional expression, and overall audio quality.
When you select Improve Delivery, HeyGen Studio opens advanced options to make your script sound more natural and engaging.
Enhance Your Voice with Improve Delivery
Enhanced Voice: Improves your AI voice by reducing background audio artifacts, balancing speech dynamics, and creating a more natural, realistic delivery.
Required Voice Engine: Enhanced Voice requires the ElevenLabs V3 voice engine. If your avatar is using a different engine, HeyGen Studio will display the following message:
"This requires ElevenLabs V3. Switching your voice engine can change how your voice sounds." Click Switch to ElevenLabs V3 to continue.
Enhance Voice lets you control how your avatar speaks by applying a specific delivery style, such as Casual, to your script.
Direct Voice requires the Panda voice engine. When enabled, the selected delivery style is applied to all scenes that use the same avatar in the current project.
How to Revert Voice Changes
If you test a new voice engine or an "Improve Delivery" feature and prefer your prior setup, you can roll back immediately without losing your settings.
After modifying an engine or delivery option, click the Revert button in the Voice panel.
Studio will restore your avatar to its exact prior voice engine and settings, skipping the need to manually re-select your old configuration.
Frequently Asked Questions & Edge Cases
What Happens When My Video Has Multiple Avatars?
Voice Settings in HeyGen Studio are configured for each individual avatar, not for the entire video.
If Scene 1 and Scene 2 both use Avatar A (for example, Leo), changing Leo's AI voice engine updates every scene where Leo appears in the current project.
If Scene 2 also includes Avatar B (for example, Sata) using a different AI voice engine (such as Fish), changing Leo's voice engine does not affect Sata. Each avatar maintains its own Voice Settings and AI voice engine independently.
How Do Scene Prompts Affect Speech Quality?
For complex script delivery, use precise scene instructions. Check out our Video Agent Prompting Guide to learn how punctuation, phonetic spellings, and context cues improve AI voice naturalness.
Credit Usage & Previews
Credit Impact
Adjusting Settings & Previews: Browsing voice settings and listening to quick audio previews does not consume extra generation credits.
Engine Switching & Exporting: Video rendering credits are calculated upon final export based on total generated video length.