Latest AI Model Releases for Image, Video and 3D (November 2026), and Where to Try Them

Short answer: In October 2026 the biggest new creative AI models were Kling VIDEO 4.0 for video, and FLUX 3 Image and Nano Banana 2.1 for images. Alongside them came new models for editable vector art, avatar video, upscaling and voice. They are all available in the Creative AI OS Model Library, so you can try them side by side in one workspace without opening new accounts. We update this page every month.

Last updated: November 4, 2026. Covers releases from October 2026.

Key takeaways

• Video is consolidating. Kling VIDEO 4.0 puts text-to-video, image-to-video, reference-driven generation and video editing into one model, with clips up to 30 seconds.
• Images are getting more controllable. FLUX 3 Image composes from up to 10 reference images and renders natively at up to 4K. Nano Banana 2.1 follows complex edit instructions more closely.
• Production formats are filling in. Editable SVG vector art, lip-synced avatar video and video upscaling cover tasks that used to need separate tools.
• You don't need to track every release yourself. Describe the result you want and Nyx can pick and run a suitable model for the job.

Video: new models this month

Kling VIDEO 4.0

Best for: trailers, longer narrative shots and storyboard-driven sequences.

Kling VIDEO 4.0 is Kling AI's new flagship. Kling 3.0 came as separate Standard, Pro, Omni and Turbo models. Version 4.0 brings text-to-video, image-to-video, reference-driven generation and prompt-based editing of an existing clip into a single model. It generates 720p or 1080p clips from 3 to 30 seconds in a single pass, with two-channel stereo audio.

What makes it useful in production is control. You can place up to 10 keyframes on the timeline, including first, middle and last frames. You can also combine up to 15 inputs in one request: keyframes, reference images, reference videos and reusable elements. That helps a character, product or location stay consistent across a whole shot. Kling has announced 4K, 10-bit HDR and a faster 4.0 Flash version as coming soon. We'll add them here once they are available.

• Game studios: cinematic trailers and in-game cutscene drafts built from approved key art.
• Marketing teams: 30-second product films directed shot by shot from a storyboard.
• Less suited to: fast, low-cost concept exploration. A lighter model is usually a better first pass.

Grok Imagine Video 1.5 Lite

Best for: fast, low-cost video drafts with native audio.

A lower-cost tier of Grok Imagine Video 1.5, with clips up to 15 seconds. It suits early concepts and social variations, before you commit to a flagship model for the final cut.

MiniMax H3 Lite and H3 Highspeed

Best for: high-volume, low-resolution video drafts.

MiniMax added two lighter tiers to its H3 family. They take text, image, video and audio inputs, which makes them practical for testing many ideas quickly.

P-Video-2-Pro-Avatar

Best for: talking-head explainers, localized spokesperson clips and character dialogue.

It creates lip-synced talking video from one portrait and an audio track or script, up to 3 minutes long and up to 4K. Only use portraits you have the rights to.

Video upscalers: P-Video-Upscale and FLUX 3 Upscaler

Best for: finishing drafts for delivery.

P-Video-Upscale enlarges video to 720p, 1080p, 2K or 4K and keeps the aspect ratio and audio. Black Forest Labs' FLUX 3 Upscaler offers a precise mode and a creative mode.

Image: new models this month

FLUX 3 Image

Best for: key art, product scenes and layouts with a lot of text.

FLUX 3 Image is Black Forest Labs' first image model built on its multimodal FLUX 3 backbone. It generates natively at up to 4K and composes from up to 10 reference images while keeping identities and fine details. You can place elements with bounding boxes and edit text in place. For campaign work, that means a character, an outfit and a location can come from separate references and still end up in one consistent scene.

Two open-weights versions followed: FLUX 3 Image Klein, a distilled version built for speed and fast iteration, and FLUX 3 Image Klein Base, the full model for fine-tuning.

Nano Banana 2.1

Best for: precise multi-step edits, recurring characters and products, and layout-heavy design.

Google's update to Nano Banana 2 improves graphic composition and subject consistency, and follows complex prompts and edit instructions more closely. It works from text alone or from up to 14 reference images plus a reference video. It renders at 1K, 2K and 4K, with three thinking levels that trade speed for reasoning depth.

Ideogram Vector 1.0

Best for: icons, UI elements, logos for mock-ups and scalable graphics.

It creates editable SVG artwork from a prompt or from an existing image. Game UI and marketing teams get vector files they can keep editing, not flat pixels.

Qwen-Image-2.1 Pro and Turbo, and Bria FIBO 2.0

Best for: structured, repeatable image generation and instruction-based edits.

Alibaba's Qwen-Image-2.1 comes in Pro and lower-cost Turbo tiers, both with multi-reference editing. Bria FIBO 2.0 generates from structured prompts in aspect ratios from 5:1 to 1:5, and FIBO 2.0 Edit handles edits with masks and up to four reference images.

Audio and 3D

Audio: HeyGen Voice 1.0 and the faster Voice 1.0 Turbo add text-to-speech with public voices and instant voice cloning from a single recording. They are useful for voice-overs, localization and scratch dialogue. Only clone voices you have permission to use.

3D: no major new 3D generation model arrived in October. The current image-to-3D and text-to-3D models remain available in the Model Library. For how to choose between them, read our guide to Blender with Python vs AI 3D generators.

Comparison table: October 2026 releases

The Model Library, with 325+ models, is included on every plan except Creator. See pricing for details, or browse our practical list of AI models.

How to pick the right model

New models arrive every month, and the best choice depends on the job, not on which model is newest. A practical approach:

1. Start from the result. For example, a 20-second trailer from three key-art frames, or 40 on-style item icons.
2. Draft with a fast model, finish with a flagship. Explore ideas cheaply, then render the final version at full quality.
3. Compare on your own material. Run the same prompt and references through two or three models in one workspace and judge the results side by side.
4. Let Nyx do the routing. In the Creative AI OS, you describe what you want and Nyx picks and runs suitable models to deliver outputs ready for review.

Start Creating and try this month's models in one workspace.

One API for studios and custom agents

Studios that build their own tools or agents can reach the same Model Library through one API, with access, usage and model choice kept under the studio's control. When a new model arrives, it becomes available without a new vendor contract or integration. See Integrate for how it works.

A note on licensing

Third-party models in the Model Library follow their own licence terms. Check them before using outputs commercially, and only use references, faces and voices you have the rights to. Output from Virtuall's own approved models is commercially safe.

FAQ

What are the latest AI video models in November 2026?

The most notable recent release is Kling VIDEO 4.0. It generates 1080p clips up to 30 seconds with stereo audio, up to 10 keyframes and up to 15 combined references, and it can also edit existing video. Lighter options released in October include Grok Imagine Video 1.5 Lite and MiniMax H3 Lite.

Where can I try Kling 4?

Kling VIDEO 4.0 is available in the Creative AI OS Model Library on Pro and higher plans, alongside other leading video models, so you can compare results in one workspace.

What is new in FLUX 3 Image?

FLUX 3 Image generates natively at up to 4K, composes from up to 10 reference images, places elements with bounding boxes and edits text in place. Open-weights Klein and Klein Base versions are also available.

How often is this page updated?

Every month. We add the month's notable new image, video, audio and 3D models and update the date at the top.

Do I need a separate account for each AI model?

No. The Creative AI OS gives you access to 325+ models from one workspace and one plan. Studios can also reach them through one API.