All posts
Model Launches

Avatar X on Flixly: Talking-Avatar Video That Keeps Your Identity

Avatar X is Mirage's most advanced avatar model — type a script for a lifelike stock presenter, or drive any face from a photo or clip with your own voice track, up to 3 minutes in one take. It's live on Flixly.

August 11, 2026

TL;DR

Avatar X is Mirage's newest avatar model, built for identity preservation and expressive delivery. In Text to Video you write a script (50-1,500 characters) and a stock presenter speaks it. In Reference to Video you upload a voice track of 3 seconds to 3 minutes and drive any identity — a photo, a short clip, or a stock avatar. Both modes are live on Flixly today, billed by the second of generated video.

Avatar X is Mirage's most advanced avatar model, and it is live on Flixly in both of its modes: script-driven text-to-video and audio-driven reference-to-video.

What makes Avatar X different

  • Identity preservation. The face you provide — a photo, a clip, or a stock presenter — stays consistent frame after frame. No drifting features, no uncanny morphing halfway through a take.
  • Expressivity. Delivery follows the words: emphasis, pauses, gesture, and facial emotion track the script or the audio rather than looping a canned animation.
  • Long takes. A single generation runs up to 3 minutes of speech — enough for a full product pitch, an explainer, or a UGC-style testimonial without stitching clips.
  • Two ways in. Type a script and let a stock presenter deliver it, or upload your own voice track and drive any identity with it.

The two modes

Text to Video — write a script. Pick one of 24 stock presenters (portrait for vertical platforms, "(16:9)" variants for landscape, plus four news-desk anchors) and write 50 to 1,500 characters. The presenter speaks it in a matching voice. Every 50 characters is roughly four seconds of speech.

Reference to Video — bring a voice. Upload the audio your avatar should speak — 3 seconds to 3 minutes — and give it an identity: a single photo (9:16 or 16:9, at least 512px on the short edge) or a short clip (1-60 seconds). Or attach audio alone and pick a stock presenter to deliver it. Lips, timing, and expression follow your track.

Specs

Property Value
Modes on Flixly Text-to-Video, Reference-to-Video
Script length 50-1,500 characters (~4s per 50 chars)
Driving audio 3 seconds - 3 minutes (mp3, wav, m4a, aac, ogg)
Identity reference one photo OR one clip (1-60s), 9:16 or 16:9
Stock presenters 24 — portrait, landscape, and news-desk variants
Aspect ratios 9:16 (default), 16:9
Output MP4 video with voice

How to use it

  1. Open the Video Generator and pick the tab you need — Text or Reference.
  2. Pick Avatar X from the model selector.
  3. In Text mode: choose a presenter and write your script the way it should be spoken — punctuation shapes the pacing, short sentences land harder on camera.
  4. In Reference mode: attach your voice track, then add a photo or clip for the identity (or leave it to a stock presenter). Clean, dry audio without background music gives the best lip-sync.
  5. Generate. The clip length follows your script or audio automatically.

Pricing notes

Avatar X is billed by the second of generated video, and the length follows your input — there is no duration picker. A 30-second voice track costs half of a one-minute one; a 500-character script costs roughly a third of a 1,500-character one. The estimate on the page always shows the exact credit cost for your current script or audio before you generate.

Try it

Avatar X is live in Text to Video and Reference to Video.

Frequently Asked Questions

What is Avatar X?

Avatar X is Mirage's most advanced generation model, built for talking-avatar video. Its headline strengths are identity preservation — the face you provide stays that face, frame after frame — and expressivity: delivery, gesture, and emotion that follow the words rather than a loop. It runs in two modes on Flixly: script-driven (Text to Video) and audio-driven (Reference to Video).

How does the script mode work?

Write a script of 50 to 1,500 characters and pick one of 24 stock presenters — portrait or landscape variants, plus news-desk anchors. The avatar speaks your script with a matching voice. As a rule of thumb, every 50 characters is about four seconds of speech, so a full 1,500-character script runs a bit over two minutes.

Can I use my own face and voice?

Yes — that's Reference to Video. Upload the audio you want spoken (3 seconds to 3 minutes) plus an identity reference: a single photo or a short clip of up to 60 seconds. The avatar keeps that identity and speaks your audio with synced lips and natural movement. You can also drive a stock presenter with your own voice by attaching audio alone.

What are the input requirements?

Audio: mp3, wav, m4a, aac or ogg, between 3 seconds and 3 minutes. Identity photo: 9:16 or 16:9, at least 512 pixels on the short edge. Identity clip: 1 to 60 seconds, 9:16 or 16:9. Use a photo or a clip, not both — Flixly picks the clip if you attach two. Scripts run 50 to 1,500 characters.

How is Avatar X priced?

By the second of generated video. There's no duration picker — the length follows your script or your audio, and the exact credit cost is shown on the page before you generate: longer scripts and longer voice tracks cost proportionally more.

Where can I try it on Flixly?

Avatar X is live in Text to Video and Reference to Video on the Flixly dashboard. Pick it from the model selector, choose a presenter or upload your references, and generate.

Tools mentioned in this post

Avatar XMirageAI AvatarTalking HeadText to VideoReference to VideoLip SyncUGCAI VideoFlixly

Ready to create with Model Launches?

Jump straight into Flixly's AI studio and try model launches with 50+ models — free to start.