TechPrufer
Get started
DeepMind
DeepMind logoDeepMind · Feature trail

Generative Media Models (Image and Video)

14 posts · Jun 2025 – Sep 2026

What this trail covers

Gemini models are enhanced for live voice interactions, including improved speech-to-text and speech-to-speech translation. Gemini 3.8 Live and 3.8 Live Extended Thinking offer advancements in near real-time reasoning for voice agents, with improved conversational intelligence, fluid dialogue, visual grounding, and multi-step reasoning. These models support over 97 languages mid-conversation, execute tools and API calls in the background, and provide live progress narration for complex tasks. Nano Banana Pro, built on Gemini 3 Pro, offers state-of-the-art image generation and editing with enhanced reasoning, world knowledge, and real-time information integration. It excels at generating accurate, context-rich visuals, including educational explainers, infographics, and diagrams. It also produces images with legible text in multiple languages, supports blending up to 14 images while maintaining consistency of up to 5 people, and offers studio-quality creative controls like localized editing, camera angle adjustments, focus changes, color grading, and scene lighting transformations. It supports aspect ratios and resolutions up to 4K. The technology is integrated across consumer, professional, developer, and enterprise products, including the Gemini app, Google Search, NotebookLM, Google Ads, Google Slides, Vids, Gemini API, Google AI Studio, Google Antigravity, Vertex AI, Gemini Enterprise, Flow, and ProducerAI. All media generated by Google's tools are embedded with SynthID digital watermarks for verification. The Gemini app allows users to upload images and ask if they were generated by Google AI. Free and Google AI Pro tier users will have a visible watermark (Gemini sparkle), while Google AI Ultra subscribers and Google AI Studio users will not have a visible watermark.

All DeepMind trails
  1. Introducing agentic video understanding with GeminiIntroduces agentic video understanding across Gemini 3.7 Flash, 3.6 Flash, and 3.5 Flash-Lite models. This capability allows Gemini to dynamically search, scan, and inspect video segments using native video tools, improving performance and reducing token usage and costs. Key features include sub-second moment retrieval, more accurate anomaly detection, and precise counting. Benchmarks show up to 66% cost reduction and 88% token reduction with up to 7% accuracy improvement, particularly for long-form video. The system works by enabling Gemini to actively decide what to watch, at what speed, and through which modality, invoking internal tools to load relevant video parts.
  2. Gemini Omni 1.1 Flash lets you build with more controlGemini Omni 1.1 Flash introduces enhanced creative controls for generative video. Scene extension now analyzes up to 10 seconds of prior context, allowing for cumulative video lengths up to 40 seconds. Users can specify first and last frames for precise control over transitions. Efficient 360p preview generation is available for rapid prototyping, and upscaling to 1080p and 4K resolution is supported. Video references of up to three seconds can be used in multimodal input.
  3. Start building with Nano Banana 2 Lite and Gemini Omni FlashIntroduces Nano Banana 2 Lite, a faster and more cost-efficient image generation model (text-to-image in 4 seconds, $0.034 per 1K image), and Gemini Omni Flash, a model for video generation and conversational editing ($0.10 per second of video output). Nano Banana 2 Lite is positioned as a replacement for gemini-2.5-flash-image. Gemini Omni Flash supports conversational video editing, multimodal referencing, and real-world knowledge integration. Limitations include 10-second video generations, and incomplete support for audio/video references and character consistency. The post highlights chaining these models for end-to-end multimedia experiences and provides demo apps.
  4. Making it easier to understand how content was created and editedThis post details the scaling of SynthID, DeepMind's digital watermarking technology, to watermark over 100 billion images/videos and 60,000 years of audio across generative media models and products. It also highlights the expansion of C2PA Content Credentials integration into Pixel phones for image and video capture, and the rollout of SynthID and C2PA verification features in the Gemini app, Search, and Chrome. The post announces the launch of an AI Content Detection API on Google Cloud's Gemini Enterprise Agent Platform and continued industry partnerships to broaden SynthID adoption and advocate for provenance standards.
  5. Nano Banana 2: Combining Pro capabilities with lightning-fast speedIntroduces Nano Banana 2 (Gemini 3.1 Flash Image), a new state-of-the-art image model that combines advanced intelligence and visual quality with lightning-fast speed. Key improvements include enhanced subject consistency (up to five characters, 14 objects), precise instruction following, production-ready aspect ratios and resolutions (512px to 4K), and visual fidelity upgrades (vibrant lighting, richer textures, sharper details). It leverages Gemini's real-world knowledge and real-time web search information for more accurate rendering and enables precision text rendering and translation within images. The model is integrated into various Google products and enhances provenance with SynthID and C2PA Content Credentials.
  7. Veo 3.1 Ingredients to Video: More consistency, creativity and controlVeo 3.1 introduces significant improvements to the 'Ingredients to Video' capability, enhancing expressiveness, creativity, and control. Key technical advancements include: improved identity consistency for characters across scenes, enhanced background and object consistency, and the ability to seamlessly blend disparate elements (characters, objects, textures, backgrounds). The update also introduces native vertical outputs (9:16 aspect ratio) for mobile-first video creation and state-of-the-art upscaling to 1080p and 4K resolution for high-fidelity production workflows. The post also mentions the integration of SynthID digital watermarking for AI-generated videos and expanded verification tools in the Gemini app.
  8. Build with Nano Banana Pro, our Gemini 3 Pro Image modelThis post introduces Nano Banana Pro (Gemini 3 Pro Image), a new image generation model. Key technical contributions include: higher fidelity image generation, improved text rendering accuracy, integration of Google Search grounding for data-driven outputs, enhanced controls for physics (lighting, camera, focus, color grading) and composition, support for 2K and 4K resolutions, consistent resemblance for up to five individuals, and the ability to blend up to fourteen standard inputs. It also features advanced image-to-image generation for localization and semantic context changes, and direct integration of SynthID digital watermarks for provenance.
  9. Introducing Nano Banana ProIntroduces Nano Banana Pro, a new state-of-the-art image generation and editing model built on Gemini 3 Pro. Key technical advancements include: enhanced reasoning and world knowledge for generating context-rich visuals and infographics; improved text rendering within images, supporting multiple languages and various fonts/calligraphy; increased consistency by design, allowing blending of up to 14 images and maintaining resemblance of up to 5 people; and upgraded creative controls for localized editing, camera angle adjustments, focus changes, color grading, and scene lighting transformations. The post also details the integration of SynthID for AI-generated image verification and the rollout strategy across various product tiers.
  10. Behind “ANCESTRA”: combining Veo with live-action filmmakingIntroduced new Veo capabilities for personalized video generation (fine-tuning Imagen and using Gemini for prompts, then Veo's image-to-video), precise motion matching (tracking draft camera motion from 3D models and guiding with text prompts), and blending generative video with live-action footage (using 'add object' to insert generated babies into live scenes, and compositing multiple generated elements with traditional VFX).