AI Image and Video Release Notes

Latest AI Image and Video Updates

• Aug 29, 2026

  • Date parsed from source:
    Aug 29, 2026
  • First seen by Releasebot:
    Aug 29, 2026
  

  Midjourney

  Edit Image Quality Update

  Midjourney updates its V8.2 edit model for better image quality.

  Hey everyone, we've updated our V8.2 edit model for better image quality. If you've had any issues in the last 24 hours give it a new shot and keep giving us feedback. Thanks! More updates soon.

  Original source
• Aug 29, 2026

  • Date parsed from source:
    Aug 29, 2026
  • First seen by Releasebot:
    Aug 29, 2026
  

  Civitai

  Weekly Update: August 22 to August 29, 2026

  Civitai expands its changelog with Hubs for custom feeds, a Sticker Book, new video and image editing options, flatter licensing fees, reusable text blurbs, draft vs published image views, remix provenance, and a broad set of quality-of-life and bug fixes across the platform.

  Civitai Changelog: August 22 to August 29, 2026

  🚀 New Features

  Hubs

  • Build your own feed. Hubs are open to everyone now. Pick the creators, models and collections you want and a Hub pulls them into one image feed. You can add a model or a creator to a Hub straight from its own page.
  • Share, follow and duplicate. Make a Hub public and anyone with the link can open and follow it, or duplicate it as a starting point for their own. A shared link gets a preview card. Making a Hub private again closes every link you have already shared.
  • Set the content level per Hub. Each Hub carries its own rating cap, and what you see is that cap crossed with your own browsing settings, so a Hub can never show you more than you have asked for.

  Stickers

  • The Sticker Book. A new tab on your profile collecting every sticker you have been given and every one you have placed.
  • Search and sort the placement tray, so a large collection is usable.
  • You hear back on a placement. You are told when one is declined, when it expires, and when a moderator takes it down.

  Generation

  • Wan Video 3.0. Alibaba's newest video model, in the generator now. 480p, 720p and 1080p, clips from 2 to 30 seconds, text to video and image to video. On image to video it takes the aspect from your source image rather than asking.
  • Krea 2 image editing. The edit workflow is now selectable in the picker, running on the Raw and Turbo builds with Turbo as the default. Community LoRAs stack on top of it.
  • Veo 3.1. Veo has moved off the retired 3.0 endpoints.

  Monetization

  • The per tier licensing fee caps are gone. There is one flat ceiling now, and any tier can raise a price up to it whenever the creator chooses. Nothing rose on its own: prices that were sitting above the old caps were written down to what buyers were actually being charged first, so the change costs no existing buyer anything and raising a price is a decision you make.
  • The eligibility floor is shown, rather than only appearing when a save is refused.

  Additional New Features

  • Reusable text blurbs. Write a piece of text once, name it, and drop it into any supported editor by reference. Editing the blurb updates every page it appears on without you touching those pages. It works in model and version descriptions, articles, bounties and Cosmetic Shop items.
  • A Published and Draft toggle on your image feed. Draft images had no home in the product, and the only thing surfacing them was a bug that fused your drafts, bounty entry uploads and orphans into every feed whether you wanted them or not. Drafts now have their own view and stay out of the rest.
  • See what an image was remixed from. Image detail shows verified provenance for the source it came from.
  • Remixed images are discoverable in the feed.
  • Filter your Models list as you type in Creator Studio.
  • Model file details come from the file itself. Precision and type are read out of the tensor header rather than guessed.
  • Training runs read their status and epochs from the orchestrator, untrained epochs are filtered out of results, and epoch filenames drop the version name while keeping the id.
  • Hash and AIR follow the download variant you picked.

  🔄 Changes

  • Reward and purchase multipliers end when a subscription actually expires. A membership whose billing period had ended kept paying its multiplier indefinitely, in some cases for months.

  ✨ Quality of Life

  • Restricted resource messaging says the same thing everywhere.
  • The collections sidebar got its collapse toggle back, and stays usable when your browsing settings filter what it can show.
  • The detail modal loads the whole post, not just the slice the feed had already fetched.
  • A sale badge shows the deepest sale on the card.
  • Fee editors start from the suggestion for that model type.
  • The account switcher keeps the signed-in user's entry intact and refreshes an avatar when it changes.

  🐛 Bug Fixes

  • The tag chip row shrinks instead of shoving the filters off the modal.
  • A LoRA could inherit a checkpoint's per image fee.
  • Some images lost all their metadata. An AddNet image with an unreadable weight dropped everything rather than the one bad value, and SwarmUI versions were read from the wrong key.
  • Resources found only by hash are now reported as unmatched rather than passing silently.
  • Search sorted on attributes the model index doesn't expose.
  • Removed comments were still counted, and could still appear in a notification preview.
  • Feed filters. The Scheduled filter returned every unpublished image you own, drafts were served to a creator who asked for scheduled, a paged fallback restarted the feed at page one, and a sort you can't pick was still the one the feed ran.
  • A released version's price can be raised again.
  • An ownership check could authorize a linked resource's model.
  • The moderation card shows on models pending publish, and "Use as cover image" is gated properly and fails loudly when refused.
  • A perkless Buzz grant stripped a paid member's multiplier.
  • A telemetry outage could fail an upload, and a failed cache clear could error a write that had already gone through.
  • Challenges. A finished challenge's edit page opens read only, the daily theme is anchored to the featured resource, and one failed group no longer abandons a whole judging round.
  • Email domain checks are enforced on every path that writes an account's email.
  • The by-hash batch API built a download URL from a header that doesn't exist.
  • The collections showcase confirm dialog closes when you press No.
  • The guided tours had four faults, all live for months. The remix button opens a menu rather than going straight to the generator, so the tours that point at it now cover the menu too. Alongside that: two model page steps were silently skipping themselves, the tooltip's close button did nothing in every tour (Escape still worked), the submit step had been pushed out of the signed out generator tour, and one step scrolled hundreds of images past the heading it was describing.

  🧪 In Testing

  Available to a limited group while we finish testing. If you can't see these yet, that's expected.

  • Scheduled model sales. Set a sale to start and end on a schedule.
  • Creator announcements. Post your own announcement from the Creator Studio composer.
  • The DM redesign, Seven more membership chat themes.
  • More ways into the remix gallery, from the generator, the image menu and the post editor.

  📣 Official Posts

  • Licensing Fees: The Cap Is Gone - August 26
  • Sticker Book: Every Sticker You've Given and Got - August 27
  • Hubs: Build a Feed From Your Favorites - August 27

  Original source

  Show more
• All of your release notes in one feed

  Join Releasebot and get updates from Midjourney and hundreds of other software products.

  Create account

  Get updates with: RSS Email CLI MCP API Slack n8n
• Aug 27, 2026

  • Date parsed from source:
    Aug 27, 2026
  • First seen by Releasebot:
    Aug 28, 2026
  

  Midjourney

  Edit Model for V8

  Midjourney releases its first V8.2 image edit model, bringing image-to-image editing, inpainting and outpainting, up to four reference images, and support for personalization, moodboards, and srefs. Updated web UIs are also included as testing expands.

  Hey everyone! Today we’re gonna start letting everyone test our first V8.2 image edit model!

  This model supports:

  1. Editing images with instructions
  2. Generating images with other images (replacing omni-reference) with up to 4 image references at once
  3. Changing specific areas of your images (inpainting) or expand your canvas (outpainting)
  4. Using personalization, moodboards and srefs (moodboards/srefs may work best with some extra prompt direction in this version)

  Ways to use the model:

  • Drag images into the prompt bar under "attach to prompt”
  • Click ‘edit’ in the lower right in the lightbox
  • Click the ‘edit’ tab on the left and upload an image
  • Type --edit url... on Discord

  We’ve updated UIs on both the main midjourney.com website and alpha.midjourney.com, alpha will be changing rapidly with lots of experiments on how to remake the interfaces for this model.

  There’s gonna be a lot of edge cases here so we’ll need your help finding situations that don’t work as expected. If an image doesn’t work well try clicking the sad face emoji on the right side of the lightbox or give us the jobid/url for the image while it's opened in the lightbox.

  We’d also love your thoughts on how we can improve our user interfaces for this model from here. Drop your thoughts or model edge-cases in #ideas-and-features.

  If you want to share what you’re doing with the new model, feel free to chat or share in #edit-showcase.

  There will be a lot of fast-follow updates from here, but we’re ready to start testing as a larger community and hope you have fun!

  ❤️

  Original source

  Show more
• Aug 26, 2026

  • Date parsed from source:
    Aug 26, 2026
  • First seen by Releasebot:
    Aug 28, 2026
  

  Runway API by Runway AI

  WAN 3.0 on Runway Dev

  Runway API adds WAN 3.0 for text and image to video with native audio and reference-driven shots.

  August 26th, 2026 - WAN 3.0 (wan3) is now available on Runway Dev.

  Generate up to 30 seconds of video with native audio from a text prompt or a start-frame image.

  WAN 3.0 is built for reference-driven shots — pass images, video, or audio to keep characters and style consistent, or use first and last keyframes to lock the edit.

  Output at 480p, 720p, or 1080p.

  Billed at 5 credits per second at 480p, 10 at 720p, and 20 at 1080p.

  Use the text to video and image to video endpoints to get started.

  Original source

  Show more
• Aug 26, 2026

  • Date parsed from source:
    Aug 26, 2026
  • First seen by Releasebot:
    Aug 26, 2026
  

  Runway API by Runway AI

  Muse Image on Runway Dev

  Runway API adds Meta Muse Image for text-to-image generation with reference image editing on Runway Dev.

  August 26th, 2026 - Meta Muse Image (muse_image) is now available on Runway Dev. Generate images from text with optional reference images — when references are provided, the model edits and combines them as the prompt describes. Supports prompts up to 4,000 characters, up to 10 reference images, outputCount from 1–10, and eight width:height ratios plus auto. Billed at 1 credit per image, multiplied by outputCount. Use the text to image endpoint to get started.

  Original source
• Aug 24, 2026

  • Date parsed from source:
    Aug 24, 2026
  • First seen by Releasebot:
    Aug 28, 2026
  

  Runway AI

  Latest updates

  Runway AI adds org setting alerts for missing SSO enforcement and low-credit email setup with direct fix links.

  Enterprise

  Org Config Alerts

  Org admins now see alerts in org settings when workspaces are missing SSO enforcement or lack a low-credit email, with a link to fix each issue directly.

  Original source
• Aug 24, 2026

  • Date parsed from source:
    Aug 24, 2026
  • First seen by Releasebot:
    Aug 28, 2026
  

  Runway AI

  Latest updates

  Runway AI adds Enterprise Default SSO for new workspaces, automatically applying org SSO settings to each new workspace.

  Enterprise

  Default SSO for New Workspaces

  Org admins can now set a default SSO configuration that automatically applies to every new workspace created in the org. This removes the need to manually copy SSO settings each time a new workspace is spun up.

  Original source
• Aug 24, 2026

  • Date parsed from source:
    Aug 24, 2026
  • First seen by Releasebot:
    Aug 28, 2026
  

  Runway AI

  Latest updates

  Runway AI adds Wan 3.0 in tool mode and workflows for near state-of-the-art video with audio and multi-reference inputs.

  Paid Plans

  Wan 3.0

  Wan 3.0 is now available in Runway in tool mode and workflows. Generate near state-of-the-art video with audio, using multiple image, video, and audio reference inputs, at 480p, 720p, or 1080p.

  Original source
• August 2026

  • No date parsed from source.
  • First seen by Releasebot:
    Aug 22, 2026
  

  Synthesia

  Keep layouts looking polished for your translated videos

  Synthesia improves video translation to keep on-screen text from overflowing. Translations now adapt to each text box, and SyRen scales long text down so translated slides stay polished and closer to the original layout.

  Translate videos without text overflow

  Translations adapt to each text box, and long strings scale down to keep your slides looking polished.

  Translating a video into a new language has always come with a catch: the translated text often doesn’t fit the layout it lands in. That means manually fixing text elements scene by scene, for every single language.

  We’ve optimized our translation workflow to better fit the space each text element has on your scene, so your scenes are more likely to stay true to your original layout.

  How it works and why it matters

  When you translate a video, on-screen text can expand in some languages and overflow its layout. That’s one of the biggest sources of manual work in a translated video—and it’s exactly what we’ve been working to fix.

  What’s new:

  • Translations take each text box’s available space into account, so the translation engine generates text sized to fit the existing box (not just a literal, line-by-line translation)
  • If a translation still runs long, SyRen (our AI rendering model for creating video scenes) proportionally scales the text down so it fits without breaking your layout

  Why you’ll care:

  • Your translated videos are more likely to look ready to share without slide-by-slide fixes
  • Less manual touch-up per language makes it faster (and cheaper) to add more locales

  For more on translating videos, check out our documentation.

  Original source

  Show more
• Aug 21, 2026

  • Date parsed from source:
    Aug 21, 2026
  • First seen by Releasebot:
    Aug 22, 2026
  

  Synthesia

  Dub multiple videos at once with bulk dubbing uploads

  Synthesia adds bulk dubbing upload for up to 10 videos at once, letting users set languages and dubbing options once, then process each file in the background with per-video status, source language detection, and one batch credit estimate.

  Upload and dub up to 10 videos in one go—set languages and options once, then let every file process in the background.

  You can now dub multiple videos in one go. Upload up to 10 files, choose your settings once, and we’ll dub each video in the background.

  What’s new

  If you’re localizing a library of content, the old flow meant repeating the same steps video after video. Bulk Dubbing Upload turns that into one batch—so you can stay consistent and save time.

  Upload a batch

  • Drag and drop up to 10 videos on the Dubbing page, or select them from your computer
  • See a thumbnail, upload progress, and status for each file
  • Rename or remove videos before you submit

  Set your dubbing settings once

  • Choose your target languages
  • Pick full dub with lip-sync or audio-only dub
  • Set lip-sync and video timing options for the whole batch

  We handle the rest per video

  • Source language detection runs automatically for each file
  • If detection fails for a file, you can set the source language for that video
  • You’ll see one credit estimate for the full batch before you submit

  For more on how it works, check out our documentation.

  Original source

  Show more