OmniHuman 2.0: The Voice Sounds Sad. Why Is the Face Smiling?

An experiment on audio-driven avatar animation examines how portrait expression, voice delivery, and text prompts influence visible emotion, while clarifying that all completed examples used Avatar Omni Human 1.5 rather than Omni Human 2.0 because a 2.0 request failed in the Atlas test environment. Using the same smiling portrait and sad 9.68-second voice recording, two videos prompted for restrained sadness and sustained cheerfulness produced differing gaze, mouth, and gesture choices, although both retained the source smile initially and neither followed every prompt detail exactly. A separate anger-versus-grief comparison changed the portrait, voice delivery, and direction together, producing visibly distinct outcomes but not isolating any single cause. The account argues that voice synchronization and emotional expression should be evaluated separately, with clips reviewed both muted and with audio at the same timestamps, and recommends changing only one input at a time when troubleshooting. It also notes Omni Human 2.0’s listed image-and-audio workflow and specifications, cautions against treating 1.5 outputs as 2.0 performance evidence, and advises checking model availability, completed outputs, rights, consent, and disclosure requirements before publishing synthetic animated footage.

No tracked trend matches for this post yet.

Use this post, company, and trend context to find content marketing opportunities, perform competitive analysis, or address product feature gaps via the Plushcap MCP server or the Plushcap API.