Veo 3 is one of the best AI video generation models available today, if not the best. However, many users are already expecting its replacement – the Veo 4, which should fix any problems and allow us to create truly realistic, complex scenes using AI.
In this article, we’re going to speculate when Veo 4 comes out, and what it will bring when it does. Let’s get stuck in.
Veo is Google DeepMind's family of AI video generation models. The series includes Veo (announced May 2024), Veo 2 (released December 2024), and Veo 3 (released May 2025).
Google developed these models to compete with OpenAI's Sora and Runway's Gen-3. They generate videos from text prompts or images, with outputs ranging from 4 to 8 seconds depending on the version.
The models are available through Overchat AI Video Generator, Google's VideoFX platform, AI Test Kitchen, and third-party APIs like fal.ai. Each version has brought major improvements in video quality, consistency, and generation speed.
Veo 3 launched in May 2025, marking a breakthrough in AI video generation by becoming the first model to generate synchronized audio alongside video. This includes dialogue, sound effects, and ambient noise that match the visual content.
When it was first released, this model was the best there was for physics, lighting and animation. But it was later beaten by OpenAI's Sora 2.
Technical specifications:
✅ 1080p output
✅ 8-second clips
✅ 16:9 (landscape) and 9:16 (portrait) aspect ratios
✅ Native audio generation with lip-sync capability
There are two Veo 3 variants: Standard Veo 3 produces maximum quality, while Veo 3 Fast is optimized for speed.
Veo 3 is very good at following instructions about camera angles, movements, and how to compose a scene. The model understands difficult instructions like "dolly zoom" or "tracking shot" and carries them out successfully. The Fast version is quicker and cheaper, but isn't as good.
Veo 2 came out in December 2024, and Veo 3 followed in May 2025 — exactly 5 months later.
If Google maintains this pace, Veo 4 should arrive around October or November 2025. However, they might wait until December to match Veo 2's release timing.
Another possibility is May 2026 at Google I/O, where the company traditionally announces major AI products. This would give them a full year to develop significant improvements.
Competition plays a huge role in timing. OpenAI's Sora 2 already surpassed Veo 3 in quality, which might push Google to release sooner rather than later.
✅December 2025 for Veo 4 seems to be the most realistic prediction, as it gives them enough time to come out with a major leap to compete with whatever OpenAI releases next.
Now let's talk about what Veo 4 might bring to the table. Here are the most likely upgrades:
Let’s break down each of them in more detail.
Current Veo models max out at 8-second clips. Veo 4 will likely push this to 15-30 seconds, matching or exceeding Sora 2's capabilities.
This is highly likely, as NVIDIA already demonstrated 1-minute consistent videos, so the technology exists — Google just needs to implement it efficiently.
Veo 4 might let you upload your photo and voice to create personalized videos, just like Sora 2.
This feature would use your image as a reference to generate consistent character appearances throughout the video. Your avatar would move naturally and match the scene's lighting.
Voice cloning could sync with the avatar's mouth movements. You'd type dialogue, and the video would show you speaking those words perfectly.
Veo 3 sometimes struggles with keeping objects and backgrounds stable across frames. Characters might change clothing colors or furniture might shift position slightly.
Veo 4 should maintain perfect consistency throughout the entire video. This improvement comes from better temporal understanding. The model needs to remember what it generated in previous frames and stick to those decisions.
Veo 3 already understands camera movements, but sometimes the instruction following doesn’t work. For example, you might describe a shot, and Veo 3 will reverse the camera movement. It will start from the middle or last frame and work backwards. It might also miss secondary word details.
In contrast, Veo 4 should nail every instruction you give it.
Veo 4 might generate the same scene from multiple camera angles simultaneously. You'd write one prompt and get shots from the front, side, above, and behind.
This feature would work like having 5 cameras filming the same action. Google's competitors haven't cracked this yet. If Veo 4 delivers multi-angle generation, it would leap ahead of Sora, Runway, and every other video AI on the market.
Current models force you to start again if you don't like one detail, but rumour has it that Veo 4 could allow you to modify videos while they're being generated.
This feature would be absolutely revolutionary! It would not only transform how we work with AI videos, but also reduce costs.
These generations can be expensive, and we currently work through trial and error. The fewer attempts you have to make to achieve the desired result, the better.
Veo 4 will likely appear in the same places as Veo 3, plus potentially new platforms. Here's where you'll find it:
Overchat AI will offer Veo 4 video generation with the best pricing.
Google VideoFX is Google's official platform for Veo models, offering the full feature set.
AI Test Kitchen serves as Google's experimental playground. Early access and beta features typically launch here first, though with usage limits.
Some people believe that Google will integrate Veo 4 into Workspace, just as it has with Gemini. This would enable you to generate videos directly in Google Slides or Docs. While this would be great, it's unlikely to be made available to all users, as video generation is expensive and it's hard to imagine Google offering this feature to non-paying users.
Current Veo 3 pricing gives us clues about Veo 4's cost structure. On fal.ai, Veo 3 costs roughly $23 per minute of generated video at 8-second clips.
Google will most likely keep the current tiered system:
Longer videos mean higher costs. If Veo 4 generates 30-second clips, each video could cost 4x more compute power than current 8-second outputs.
But if Sora 2 undercuts Google's pricing, expect rapid adjustments. Both companies are burning cash to capture market share.
You will also be able to access Veo 4 on Overchat AI for as little as $4.99 per week.
Veo is Google DeepMind's family of video generation models that transform text and images into videos with audio. The series launched with Veo in May 2024, followed by Veo 2 in December 2024, and Veo 3 in May 2025.
Veo 4 will likely release in December 2025 or May 2026, based on Google's current release pattern. The model should bring longer videos (15-30 seconds), better consistency, clearer text, and revolutionary multi-angle generation and real-time editing.
Are you excited for Veo 4? There isn't much time left until the release, but in the meantime, you can experiment with Veo 3 on Overchat AI and start creating stunning AI videos right now!
