VideoGenReview

What each AI video model's release notes say changed

Q3 2025 in AI video release notes

Q3 2025: 31 dated entries from 9 lines, 17 July 2025 to 28 September 2025. The table lists each; the chart counts them by kind of change. As of 2026-09-25.

Sound arrives in the notes. Veo 3's preview on 17 July generates audio with the picture, Kling adds video-to-audio in August, and the Wan 2.5 preview in September does both sound and 10-second clips. The same quarter brings Runway's Act-Two and Aleph, Wan 2.2, PixVerse V5, Luma's Ray3 with HDR, Vidu Q2, Midjourney's HD mode, Firefly's smoother motion and composition references, and Veo 3 becoming generally available on both of Google's surfaces.

Q3 2025 entries by kindCounted from the table. One note can count under several kinds.Q3 2025 entries by kindNew models and snapshots88Clip length and extension22Resolution, frame rate and shape44Sound, voice and lip-sync77References and consistency33Start, end and middle frames55Editing video that exists33Camera and several shots22Motion and performance33Faster and lighter tiers99API, plans and availability44
Fig. 1 Counted from the table. One note can count under several kinds.
Entries in Q3 2025 by kind of change.
Kind of changeEntries
New models and snapshots8
Clip length and extension2
Resolution, frame rate and shape4
Sound, voice and lip-sync7
References and consistency3
Start, end and middle frames5
Editing video that exists3
Camera and several shots2
Motion and performance3
Faster and lighter tiers9
API, plans and availability4

Inclusion rule. Kinds with at least one entry this quarter; one entry can count under several. Order. The order kinds take on every page.

1Every entry

Every dated entry in Q3 2025, oldest first.
DateLineVersionWhat changed
17 July 2025Adobe Firefly VideoFirefly Video ModelMotion fidelity is improved, for smoother transitions and more natural movement
17 July 2025Adobe Firefly VideoFirefly Video ModelA reference video can pass its composition to a new generation
17 July 2025RunwayAct-TwoAct-Two tracks head, face, body and hands
17 July 2025Google VeoVeo 3Veo 3 preview arrives in the Gemini API and generates sound with the picture
17 July 2025Google VeoVeo 3Veo 3 preview models on Vertex AI can be upscaled to 1080p
25 July 2025Midjourney videoMidjourney V1 videoLooping, set start and end images, and video in the Discord bot
28 July 2025ViduNot named in the noteLip-sync from text or audio
28 July 2025WanWan 2.2Wan 2.2 text-to-video Plus names complex movement and cinematic control
29 July 2025Google VeoVeo 3Veo 3 and the Veo 3 Fast tier become generally available on Vertex AI
31 July 2025RunwayAlephAleph, a model for editing and transforming video, opens to paid plans
31 July 2025Google VeoVeo 3Image-to-video opens for the Veo 3 preview and a Veo 3 Fast preview is released
1 August 2025Kling AINot named in the noteVideo-to-audio adds sound to any Kling video, and text-to-audio arrives
5 August 2025PixVerseNot named in the noteSound effects and Fusion (reference to video) arrive
7 August 2025Midjourney videoMidjourney V1 videoAn HD mode renders video at native 720p
7 August 2025ViduNot named in the noteAn off-peak mode at lower credit cost
12 August 2025Kling AIKling 1.6Text-to-video on 1.6 gains the pro mode
12 August 2025Luma RayNot named in the noteModify with Instructions takes edits in plain language
13 August 2025Midjourney videoMidjourney V1 videoHD video opens to the Standard plan and smaller batches arrive
15 August 2025WanWan 2.2Wan 2.2 image-to-video Flash: faster, with shot control
26 August 2025ViduNot named in the noteReference video generation from one to seven images
29 August 2025PixVersePixVerse V5The V5 model is released
5 September 2025Kling AIKling 2.1Kling 2.1 takes a start and an end frame, in pro mode only
8 September 2025Google VeoVeo 3Veo 3 on Vertex AI can make 4, 6 or 8 second clips
9 September 2025Google VeoVeo 3Veo 3 goes GA in the Gemini API with more aspect ratio, resolution and seed options
11 September 2025PixVerseNot named in the noteSound effects and text-to-speech lip-sync inside text-to-video, image-to-video and first-and-last-frame calls
12 September 2025WanWan 2.2A Wan 2.2 first-and-last-frame model for transitions between two images
15 September 2025Kling AINot named in the noteLip-sync handles several people on screen and a set start time
18 September 2025Luma RayRay3Ray3 adds reasoning, native HDR EXR output and a Draft Mode
23 September 2025WanWan 2.5Wan 2.5 preview generates synchronised audio and 10 second clips
24 September 2025ViduVidu Q2Vidu Q2 Pro and Q2 Turbo for image-to-video and start-and-end frames
28 September 2025PixVerseNot named in the noteRestyle is released

Inclusion rule. Entries whose note is dated in this quarter; undated version rows are not in any quarter. Order. By date, oldest first.

2Lines with entries

3A pipeline on current models

Release notes such as “Q3 2025” matter once a pipeline runs the models; SceneMixer's guides describe how it does:

Dubbing replaces the audio of a finished clip; the picture was generated for a different line, so mouth shapes, pauses and shot length all belong to the original language. Native-language generation gives the video model the line in the target language before it renders the shot, so the performance, the pacing and the cut points are built for that language.

SceneMixer, Native-Language Dialogue vs. Dubbing in AI Video (2026): Why the Cast Should Speak the Line (read 2026-10-13)

SceneMixer writes dialogue into the script analysis, assigns each line to a shot with a speaker and a short delivery note, and the video model performs it — picture and audio generated together, so the mouth forms that sentence in the language you chose.

SceneMixer, AI Dialogue Generator (read 2026-10-13)

4Neighbouring quarters