Google just pushed a wave of upgrades that make AI video generation, live audio AI, music creation, and even UI/UX design feel a lot closer to “real production.” If you build content, prototype products, or just like making things with modern AI tools, these changes are worth your attention.
This write-up breaks down the biggest updates: Veo 3.1 Lite for faster, cheaper AI video generation in Google AI Studio, Gemini 3.1 Flash Live for more natural live audio, a smarter way to create prompts with MyPromptBuddy, easier memory import so Gemini can carry context from other AI tools, and major improvements to Google Stitch for UI design. Finally, you will see what’s new with Gemini’s music generator, including “Fast” and “Pro” modes.
Suggested media: Add a screenshot of the Google AI Studio “Video Generation” panel (showing Veo 3.1 Lite options), and one example image of an optimized prompt vs a basic prompt. An infographic comparing costs (Fast vs Lite vs higher models) would also be highly effective.
Table of Contents
- 1) Veo 3.1 Lite in Google AI Studio: Faster AI Video Generation (and cheaper)
- 2) Gemini 3.1 Flash Live: more natural and reliable live audio AI
- 3) Better prompts for AI video and images: optimize with MyPromptBuddy
- 4) Import memories into Gemini: bring context from other AI tools
- 5) Google Stitch upgrades: UI/UX design that feels like a production assistant
- 6) Create music in Gemini: Fast (30 seconds) and Pro (up to 3 minutes)
- How to use these upgrades together: a simple workflow
- FAQ
- Conclusion: the “creator stack” is getting real
1) Veo 3.1 Lite in Google AI Studio: Faster AI Video Generation (and cheaper)
The first major change is where you would expect it: Google AI Studio. Under Video generation, you can now select Veo 3.1 light (often shown as Veo 3.1 Lite).
What stands out immediately is speed. It is described as “so fast” and “a really good job,” and the workflow is straightforward:
- You access it in AI Studio (and it can also be used via API once you have access).
- You need to bring your own API key to use the model from AI Studio.
- Right now, you can generate one video at a time.
Key controls you can set in Veo 3.1 Lite
In the UI, you can tune the video output using settings like:
- Aspect ratio: 16 by 9 for landscape (like typical videos) and 9 by 16 for vertical (like YouTube Shorts).
- Duration: choose 4, 6, or 8 seconds.
- Frame rate: currently tied to a 24 FPS requirement.
- Resolution: you can select output resolution such as 720p or 1080p.
One important note: if you push for 1080p, generation takes longer than 720p. Similarly, an 8 second output takes longer than 4 or 6 seconds. That tradeoff matters if you are producing lots of variations.
Veo 3.1 Lite pricing: designed for high-volume creation
AI video is expensive when you scale. The most practical upgrade here is cost efficiency.
The breakdown shared is essentially:
- Veo 3.1 fast: around 15 cents per video
- Veo 3.1 light: around 5 cents per video
- Veo 3.1 at 720p or 1080p differences: one example given was about 8 cents at 1080p
- Veo 2 and higher models: described as “incredibly expensive,” with Veo 3 even more so
Practical takeaway: If you are building an app, running campaigns, or generating many video drafts, Veo 3.1 Lite becomes the sensible default. You can iterate faster, test more ideas, and only “upgrade” to heavier models when the result truly needs it.
Example scenario: transforming an apple seed into an apple
A simple example was used to demonstrate the model: an apple seed turning into an apple, showing a progression like:
- seed
- sprouting
- growing into a tree
- becoming an apple
The key point is that you can describe what you want, set the resolution and aspect ratio, and then iterate. The video output can also include sound, though you may choose to omit downloading sound during testing if you are simply evaluating visuals.
Availability note: The model was described as not currently available inside Gemini, but expected to appear there “in the next few days.” If your workflow is Gemini-first, this is worth watching for.
2) Gemini 3.1 Flash Live: more natural and reliable live audio AI
Next up is a shift from “offline generation” to “live interaction.” Gemini’s new 3.1 Flash Live upgrade focuses on audio AI that feels more human and behaves more consistently.
What “Live” improves
The improvements described are centered on:
- More natural audio
- More reliable behavior
- Robust reasoning and task execution
- More intuitive interaction
You can access this through Gemini’s live features. The upgrade was described as being “better” not just in one place but across the experiences where live access exists.
Where you can trigger the new live model
Two entry points were highlighted:
- Gemini’s native “go live” flow
- Gemini through a Chrome extension, using an “Ask Gemini” live option
Practical takeaway: If you rely on live AI for conversation, coaching, ideation, or real-time problem solving, Flash Live is the upgrade that should reduce awkward pauses, improve clarity, and make the interaction more dependable.
External reference: For up-to-date model and feature details, check Google’s official AI documentation: https://ai.google.dev/.
3) Better prompts for AI video and images: optimize with MyPromptBuddy
Models matter. But prompts matter too, and the difference between a basic prompt and a well-structured prompt can be dramatic for AI video generation and AI image generation.
A workflow was demonstrated using a Chrome extension called MyPromptBuddy. The goal is not to “trick” the model. It is to help you write prompts that are specific, structured, and aligned with the type of output you want.
Basic prompt vs optimized prompt
The baseline prompt example was again: an apple seed turning into an apple.
To optimize it, the extension flow described was:
- Open the MyPromptBuddy extension
- Choose the AI video option
- Paste your prompt
- Select something like high def video showing
- Click optimized prompt
What you get back is a “much better prompt” that is described as supporting multiple modes such as:
- standard prompts
- reasoning prompts
- deep research
- AI video prompts
- AI image prompts
Why optimized prompts produce better results
In the example, the optimized prompt delivered:
- More realistic visuals
- Better transitions
- A more convincing apple appearance
That aligns with what experienced creators know: if you specify style, sequencing, and clarity about what should happen frame-to-frame (or scene-to-scene), the model has a better target to follow.
Workflow tip: shortcuts for repeatable prompts
The extension also supports creating shortcuts. If you use similar prompts often, you can copy and paste them or access them quickly while working across different LLM environments.
4) Import memories into Gemini: bring context from other AI tools
Another meaningful upgrade is about long-term personalization and reducing “context loss.” Gemini now makes it easier to import memories.
How to import memories in Gemini
The flow described was:
- Go to settings in the bottom-left
- Choose import memories to Gemini
- Follow the provided instructions
The demonstration showed copying information from another AI (like ChatGPT) and then adding it into Gemini’s memory system. It also included the idea of importing chats, not just isolated facts.
Exporting data from other providers
To move the other direction (for example from ChatGPT to Gemini), the described process was:
- In ChatGPT, open settings
- Go to data controls
- Select export data
- Use the exported file to import into Gemini
Practical takeaway: If you have been using multiple AI assistants for months or years, this reduces friction and helps Gemini become more “you,” rather than starting fresh each time.
5) Google Stitch upgrades: UI/UX design that feels like a production assistant
Design tools can be hit-or-miss with AI. The Stitch upgrade described here is focused on making design tasks easier, faster, and more iterative.
Stitch was positioned as being “so good at designing things,” with the ability to:
- choose different models (for example Flash, thinking, or pro variants)
- redesign and ideate
- even “go live” to talk through design decisions
- generate not just concepts, but code and design exports
Redesigning a UI from a screenshot
The demo workflow used a screenshot of an existing website:
- Capture a screenshot of the current landing page
- In Stitch, request: “make this experience more clean”
- Optionally upload a website URL
- Upload the screenshot file
- Run the redesign task
Once generated, it was described as producing a complete redesign with multiple layout improvements. The most important feature described is iteration: you can modify the output, preview versions, and then export results.
Export and downstream design workflows
The output options included:
- exporting and copying design assets
- moving the work into tools like Figma
- generating prototypes and multiple variations
- creating mobile versions
- even producing predictive heat maps
Why this matters: UI/UX design can be a bottleneck. When AI can handle layout iteration, variation generation, and faster prototyping, it turns design into a cycle rather than a one-time hurdle.
Internal reading suggestion: If you maintain a blog around productivity and AI workflows, consider adding related guides such as “How to write UI-focused prompts” or “AI for marketing assets: video, image, and landing pages.” (Add links to your own posts here.)
6) Create music in Gemini: Fast (30 seconds) and Pro (up to 3 minutes)
The final upgrade covered is music generation inside Gemini. This is not just background noise either. The tool supports templates and lets you describe the track you want.
Fast vs Pro modes
Two modes were highlighted:
- Fast: generates about a 30 second version
- Pro: generates up to about a 3 minute version
You can choose a template or describe your target style and mood, then Gemini generates:
- sound structure
- track audio
- descriptions of the mood and style
- a confidence level for whether it matches your needs
An example rap request (style-based, not artist cloning)
An example prompt asked for a rap that sounds like it should be on Drake’s “Iceman” album. The tool reportedly refused to recreate specific content tied to a specific artist, but it still generated a piece designed to embody the mood.
The output then described the intended vibe, with something like “Toronto style hip hop,” and provided a confidence estimate. You can play the track, view subtitles (if provided), share, and download.
Practical use cases:
- soundtracks for short videos
- music for social media content
- background tracks for product demos
- rapid ideation for creators who need something fast, then refine later
In other words, this closes the gap between “visual concept” and “audio-ready content.” When you pair Veo 3.1 Lite video generation with Gemini music generation, you get a more complete creative pipeline.
How to use these upgrades together: a simple workflow
If you want a practical way to combine everything described, here is a workflow that matches the “iterate quickly” theme:
- Start with an optimized prompt for video or image using MyPromptBuddy.
- Generate drafts in Veo 3.1 Lite using the right aspect ratio (16:9 for general, 9:16 for Shorts).
- Upgrade settings only when needed: test 720p first, then move to 1080p if the concept is worth it.
- Use Gemini Live for real-time refinement: ask for alternate directions, narration ideas, or storyboard changes.
- Design UI screens in Stitch if the project needs a landing page, prototype, or redesign.
- Add music in Gemini (Fast for quick tests, Pro for longer tracks) and export for use in your video pipeline.
This is the mindset shift: treat AI models like modular tools. Video, audio, and design start to feel like parts of one production system.
FAQ
What is Veo 3.1 Lite and where do I access it?
Veo 3.1 Lite is a faster, more cost-efficient AI video generation model inside Google AI Studio. You can access it under “Video generation.” You will need your own API key to use it from AI Studio.
Can Veo 3.1 Lite generate vertical videos for YouTube Shorts?
Yes. The model supports aspect ratio settings including 9 by 16 for vertical content, alongside 16 by 9 for standard landscape videos.
Does generating 1080p or longer durations take more time?
Yes. Higher resolution like 1080p typically takes longer than 720p. Likewise, longer outputs like 8 seconds take longer than 4 or 6 seconds.
What improvements does Gemini 3.1 Flash Live provide?
Flash Live is designed to make live audio AI more natural and reliable, with stronger reasoning and task execution during live interaction.
How can I improve AI video prompts?
Use an optimizer like MyPromptBuddy to transform a basic idea into a more detailed, structured prompt. In the example, optimized prompts produced more realistic results and better transitions compared to a simple prompt.
Can Gemini import my memories from other AI services?
Yes. Gemini includes an “import memories” feature in settings, and it can also guide you through exporting data from other providers (such as ChatGPT) for migration.
What does Google Stitch do with UI/UX design?
Stitch can redesign interfaces from screenshots or URLs, generate multiple variations, and help export or build UI prototypes, including workflows that integrate with tools like Figma.
How long are the music outputs in Gemini?
Gemini music generation supports Fast mode for about 30 seconds and Pro mode for up to about 3 minutes.
Conclusion: the “creator stack” is getting real
These upgrades point in the same direction: AI is moving from isolated demos to end-to-end creative workflows. With Veo 3.1 Lite, you can iterate video faster. With Gemini 3.1 Flash Live, you can refine in real time with more natural audio. With MyPromptBuddy, you can improve your prompt quality without starting from scratch every time. And with Stitch plus Gemini music, you can build more complete projects without stitching together five separate tools.
CTA: If you try Veo 3.1 Lite, which settings are you starting with (720p vs 1080p, 4 vs 6 vs 8 seconds, 16:9 vs 9:16)? Share your results in the comments, and explore related guides on AI workflows for video, music, and UI design on your site.
Tags
- Google Gemini
- Veo 3.1 Lite
- AI video generation
- Gemini Live
- AI music generation
- Google Stitch
- MyPromptBuddy
- UI UX design AI
- prompt optimization
- memory import
This article was created from the video Google Gemini’s New AI Video Generator is INSANE 🤯 (Veo 3.1 Lite) with the help of AI.