Changelog

See what's new in Z-AI Studio.

0.5.0 — August 8, 2026

Shared character library

Characters can now be shared across the whole workspace, generated straight from the detail page, and every result lands in that character's gallery automatically.

Features

  • Admins can switch a character to shared so the whole workspace can use it; characters that are not shared show as a locked tile.
  • A generate sidebar inside the character detail page — create images and video without leaving the screen.
  • Newly generated images and video attach straight to the character's gallery.
  • Admins can manage any user's characters.
  • New video model: Lipsync 2 Pro.

Improvements

  • Opening an image or video in detail view now updates the URL, so the link is shareable and Back closes the layer you actually opened.
  • Node-canvas uses a canvas screenshot as each space's thumbnail.

Fixes

  • Some MP4s failed to report their duration because of where the metadata table sat in the file; every layout is handled now.
0.4.0 — August 4, 2026

Faster references, smoother KOL AI

Paste from the clipboard, pull references out of your Uploads library, view them full size — and KOL AI keeps your workspace when you navigate away.

Features

  • Generator: pick a reference image straight from your Uploads library.
  • Paste a clipboard image into Generator or Generate Video to use it as a reference.
  • Every KOL AI option now ships with an illustrated thumbnail.
  • New MiniMax H3 video model, plus Seedream 5 Pro for Character Sheet.
  • Character Sheet: pick a background colour for templates 3–6.

Improvements

  • KOL AI keeps its state when you navigate away and come back.
  • Click a reference image or video to view it at full size.
  • Links shared to Facebook, Zalo or Slack now render a preview card with artwork.
  • The admin page splits model visibility into its own tab.

Fixes

  • Generate Video no longer resolves the wrong model group when you use an edit variant.
0.3.0 — July 31, 2026

Audio and subtitle toolkit

Generate music from video, turn SRT scripts into narration, translate subtitles with film-script context, plus a dedicated voice library.

Features

  • Video to Music: score a clip with ElevenLabs Eleven Music.
  • Text/SRT to Audio: turn a script into speech, with narration and dubbing tabs.
  • SRT Translate: translate subtitles against the whole script instead of line by line.
  • Auto Transcribe: transcribe with Whisper or ElevenLabs Scribe and export SRT.
  • Voice Library: audition and pick voices on a screen of their own.
  • AI shortening for lines that overrun their cue.

Improvements

  • Upgraded to ElevenLabs music v2 with better tag handling.
  • Script import and translation are one flow now, with fewer clicks.

Fixes

  • Audio tools now fail loudly when a video's duration cannot be probed, instead of quietly returning a wrong result.
  • Renamed the “Voices” menu entry to “Voice Library”.
0.2.0 — July 28, 2026

Uploads library and private storage

Everything you upload now lands in a library of its own, stored in a private bucket only its owner can open.

Features

  • An Uploads screen: revisit everything you uploaded, insert it as a reference, or delete it.
  • Click an upload to open its detail view.
  • A grid density slider for Projects and Generator — choose how many cards fit per row.

Improvements

  • Uploads moved to a private R2 bucket and are served through expiring signed links instead of public URLs.
  • Every reference asset sent to an AI provider is now hosted on R2.

Fixes

  • Private uploads are no longer mirrored to the public CDN bucket.
  • Deleting a character's images now removes the matching upload records too.
0.1.0 — July 24, 2026

Characters, new video models, and a bilingual UI

The first release with a KOL character library, a Use-as-Ref flow that follows you across tools, a batch of new video models, and a full Vietnamese–English UI.

Features

  • A KOL character library: create characters, open their detail view, and manage a gallery for each.
  • Use as Ref: send any image or video into Generator or Generate Video, landing in the right reference slot.
  • New video models: Kling V3 Turbo, Kling 3.0 Motion Control, LTX 2.3 Fast, Grok Imagine Video.
  • New image models: Nano Banana Pro, Z-Image, Grok Imagine, Seedream 5 Pro.
  • Projects open as a full-screen view when you click a card.
  • The entire interface is available in Vietnamese and English.

Improvements

  • Generator's input panel sticks to the top, and results split into History and Gallery tabs.
  • Switching models in Generate Video keeps the images, video and audio you already attached.

Fixes

  • Restored the image download button in the Projects media actions.
  • Node-canvas: Prompt Concatenator no longer reads loop edges.