Mirage Launches Avatar X: High-Fidelity Digital Twin from 10-Second Video
Captions' Mirage has launched a new AI digital human model, Avatar X. It requires only 10 seconds of video footage to generate a high-fidelity personal digital twin, shifting the technical focus from mere lip-sync to identity preservation and emotional feedback, effectively addressing the distortion issues common in long videos. Multiple testers report that its naturalness significantly surpasses existing products.
Confirmed
- Avatar X dramatically lowers the training barrier, requiring only 10 seconds of video to generate a digital twin.
- The model emphasizes identity preservation, maintaining consistency in facial movements, expressions, and identity features across landscape and portrait outputs.
- It is optimized for long videos, solving the problem of facial detail drift over time that plagues traditional avatar models.
- It features more natural micro-expressions and body feedback, reacting to audio tone with actions like frowning, head tilting, and shrugging, and accurately handling non-verbal audio cues such as laughter, crying, and yawning.
Why it matters
- @mhdfaran notes that the true challenge for AI avatars is capturing subtle facial reactions between words, and Avatar X successfully captures micro-expressions like eyebrow raises and eye movements.
- @LinusEkenstam reports that his 10-second video twin significantly outperformed other models he had tried, reducing the 'impressive at first, fake soon after' flaw.
- @Scobleizer believes the model raises the bar for identity preservation and natural expressions. He suggests that if digital twins are realistic enough, creators could record once and then appear in 39 languages with their own voice, fundamentally reshaping media production.
2026-07-28 ~ 2026-07-29 · 14 related posts
Primary sources
- Mirage launches Avatar X with stronger identity preservation and video support — mhdfaran · 2026-07-28
- Mirage’s Avatar X keeps identity consistent across horizontal and vertical video — mhdfaran · 2026-07-29
- Avatar X is pitched as capturing the micro-expressions that make avatars feel alive — mhdfaran · 2026-07-29
- Avatar X adds facial and body reactions beyond basic lip-sync — mhdfaran · 2026-07-29
- Mirage Avatar X Needs Only 10s of Footage for High-Consistency AI Avatars — mhdfaran · 2026-07-29
- Mirage’s Avatar X claims a new high for identity-preserving AI avatars — Scobleizer · 2026-07-29
- Avatar X is praised for syncing facial expression with non-verbal audio — Scobleizer · 2026-07-29
- Scobleizer says AI twins could let creators appear in 39 languages after one recording — Scobleizer · 2026-07-29
- Mirage’s Avatar X claims a digital twin from 10 seconds of video — Scobleizer · 2026-07-29
- [source] Linus Ekenstam says Mirage Avatar X can clone a digital twin from 10 seconds of video — LinusEkenstam · 2026-07-29
- [source] Captions Launches Mirage Avatar X: High-Fidelity AI Twins from Just 10s of Video — dr_cintas · 2026-07-29
- Mirage's Avatar X generates high-fidelity avatars in 10 seconds, threatening existing AI video tools — socialwithaayan · 2026-07-29
2 near-duplicate retellings: mhdfaran · Scobleizer