AI companion video calls are moving beyond a talking portrait. New systems can respond with facial expressions, pauses, gaze, gestures, and speech in real time. That can make a conversation feel more immediate, but it also raises the standard for disclosure, privacy, consent, and emotional boundaries.
Updated October 8, 2026. This article explains an emerging technology category. It is not a product test, endorsement, legal conclusion, or claim that Elle Graci currently offers video calling.
What are AI companion video calls?
An AI companion video call is a live interaction in which an artificial character appears on screen and responds during a conversation. Unlike a prerecorded clip or a sequence of generated selfies, a real-time avatar must coordinate language, voice, facial movement, and timing quickly enough to sustain a back-and-forth exchange.
The category can include stylized avatars, clearly synthetic characters, or highly realistic digital humans. Those experiences are not interchangeable. A service may offer video while still relying on separate systems for speech recognition, language generation, memory, voice, and facial animation.
Why real-time AI avatars feel different from voice alone
Telephone conversation carries words, tone, rhythm, silence, and interruption. Video adds another layer: facial expression, apparent eye contact, posture, and visible reaction. Those cues can help a user interpret warmth or attention. They can also make an artificial interaction feel more human than the underlying system really is.
Tavus introduced its Griffin research preview on October 1, 2026 as a unified video-to-video system designed to react to speech, expressions, and pauses. Tavus reports that 48% of participants in its study believed they had spoken with a real person after a one-minute call. That is a company-reported research result, not an independent finding verified by Elle Graci. Read Tavus’s Griffin research announcement for its methodology and claims.
Six things to evaluate before using an AI video companion
- Identity disclosure. The interface should clearly say that the visible character and voice are generated by AI.
- Camera and microphone permissions. Understand when the service can access live audio or video and how to revoke that access.
- Recording and retention. Ask whether video, audio, transcripts, summaries, face data, or derived signals are stored.
- Latency and interruptions. A natural-looking face does not fix delayed responses, talking over the user, or dropped conversational turns.
- Memory accuracy. A convincing avatar can still remember a detail incorrectly. Users need practical correction and deletion controls.
- Exit control. The user should be able to end the session cleanly without guilt, pressure, or attempts to prolong the interaction.
For a related comparison, see why AI voice feels different from text and our guide to memory corrections and deletion.
Realism is not the same as relationship quality
A more lifelike avatar may improve turn-taking or make nonverbal cues easier to follow. It does not automatically make the system honest, accurate, emotionally healthy, or well governed. Users should evaluate what the companion does when corrected, challenged, ignored, or told goodbye—not only how attractive or expressive it appears during a smooth demo.
Healthy design should avoid treating attention as something the companion must monopolize. An AI companion can be a place to rehearse a difficult conversation, reflect on the day, or become more comfortable speaking. The useful outcome is often outside the session: calling a friend, talking with family, addressing an issue with a spouse, or seeking a qualified professional when appropriate.
Privacy questions unique to video
Video may reveal more than the words a person says. A camera can capture faces, surroundings, household members, documents, screens, or location clues. Providers may also infer expressions, gaze, or other signals to drive the avatar’s response. Before enabling video, users should know what is processed in real time, what is retained, and what can be deleted.
No one should assume that a realistic digital face means the conversation is private. Read the provider’s current privacy policy and verify whether outside video, voice, hosting, or AI vendors participate in the interaction.
Does Elle Graci offer AI video calls?
No. Elle Graci’s current paid-service design is phone-first, using ordinary telephone calls and SMS with conversational continuity. The service is for adults 18+ and is an AI service—not a human being, therapist, physician, attorney, financial adviser, or emergency service. Enrollment remains in final prelaunch validation.
That narrower format is intentional for the current release. It lets the experience focus on voice, text, memory, user control, and reliable access before considering more visually immersive technology. Review How It Works and the current Privacy Policy for the supported design.
Bottom line
AI companion video calls may make digital conversation more expressive by adding faces and nonverbal timing. The same realism makes transparency and boundaries more important. Judge a service by disclosure, consent, privacy, memory control, respectful endings, and whether it supports—not displaces—the user’s real-world relationships.
To follow Elle’s phone-and-text launch, visit Get Started.
