Avatars, Overdub, and New Subtitle Styles

Here's what's new in Voice Creator Pro v2.4.10. Everything below is live on both the desktop app and in VCP Cloud. You can also watch the update video on YouTube.

New Feature: Avatars

Avatars turns a photo into a video of that person talking. Pick a headshot and some speech, and you get a video of them saying it. The speech can be a clip from your history, an audio file, or new lines voiced with any of the TTS models in the app.

You can also make your own avatar. Choose a place, a framing, and a style, describe the person and what they're wearing, and pick from the portraits it makes. Places can be a room you describe or a photo of a real one, and any avatar you've saved can be put in them.

• 480p, 720p, or 1080p videos up to 20 seconds long
• 4K export with FlashVSR on NVIDIA RTX 30 series or newer cards with 8 GB or more
• Image models MOSS-Image, with a faster 4-step mode, FLUX.2 klein, and Z-Image Turbo

Overdub in AI Cut

AI Cut can now change what you said. Change words in the transcript and AI Cut speaks the new ones in your own voice, with lip sync so your mouth matches. When there are two or more people in the video, the transcript shows who said each word.

New Subtitle Styles

We've added 49 new subtitle styles, including Hormozi, Kelly, Ali, Kendrick, Highlighter, and Retro 3D. Captions can now slide, rise, tilt, and wiggle, and there are new word animations like Glitch, Flicker, Swipe, Rise, and Drop. Shadows can be soft, hard, or 3D, and bold subtitles in Inter, Noto Sans, Noto Serif, and Oswald now use the real bold font.

Other Improvements

• New timeline in Viral Clips, AI Cut, and Avatars with zoom, snapping, a filmstrip, and an audio waveform
• Undo and redo buttons in Viral Clips, AI Cut, Dubbing, and Avatars
• Exports keep going when you leave the screen
• Natural pauses in dubs where the original speaker paused, including translated dubs
• Find voices in the Voices panel in the Lab and use them right away
• Scripts with speakers in the Lab can become a project with a voice for each speaker
• Premium voices in Voice Search, hand-picked by us
• Numbers read correctly in all 55 languages, including phone numbers, versions, units, and dates
• Advanced settings for each audiobook segment
• NeuTTS has been removed, bad model, not worth the bytes
• Many smaller fixes, including OmniVoice saying "dot" at line breaks and a click in Qwen3 cloned lines