Google Labs

Google Labs

Google's AI and product experiments

5.0
5 reviews

1.4K followers

Google Labs is an experimental hub where Google publicly tests early-stage AI products and features. It includes projects like AI-powered search enhancements, Workspace integrations (e.g. Gmail and Docs assistants), and generative tools such as NotebookLM.
This is the 7th launch from Google Labs. View more

Google Whisk 3.0

Launched this week
Refined edits made simple with Whisk Precise Mode
Whisk’s new Precise Mode lets you generate images using your own reference uploads for sharper character, style, and scene accuracy. Powered by veo3 and refined with Gemini 2.5 Flash, it gives you control, consistency and stunning results.
Google Whisk 3.0 gallery image
Google Whisk 3.0 gallery image
Google Whisk 3.0 gallery image
Free
Launch Team

What do you think? …

Rohan Chaubey
Hunter
📌

The Google Labs team shared with me a new update. Excited to share Whisk 3.0 with the Product Hunt community! :)

This release introduces Precise Mode! This is a big step forward for creators who need tighter control over outputs.

Unlike Creative Mode, which relies on Gemini’s visual understanding to caption your images before passing them to the model, Precise Mode directly uses your uploaded reference images with a specialized version of Imagen 4 and veo3.

The result? Outputs that stick much closer to your intended character, style, and scene references.

On top of that, they have integrated Gemini 2.5 Flash into the refine flow, so you can make fast, native edits to get your visuals exactly right.

A few quick highlights:

  • Toggle “Precise Ref” in your prompt bar settings to activate.

  • Upload up to 3 context images (subject + scene) and 1 style image.

  • Works best when you pair reference images with clear text instructions.

  • Guardrails in place to block certain photorealistic categories for safety in Precise Mode (eg. minors)

Pro tip: For identity preservation, use a clean, frontal, natural color face photo.

Super excited to see what the community creates with Precise Mode. Take it for a spin and show us what you make!

Masum Parvej

@rohanrecommends How does the model balance between multiple context images if they conflict a bit?

Jaime

@rohanrecommends  @masump it depends but it can handle it and it can do better with more specific instructions

Nika

I wanted to try it, but generation failed. 😭

Jaime

@busmark_w_nika oh no, what error are you seeing? Make sure to not use a photo of a minor

Nika

@jaime_guajardo uh, I do not remember, as soon as I couldn't get the result, I closed the window. 😅

Cruise Chen

No way, Precise Mode with reference uploads is exactly what I’ve needed—finally, I can get consistent character styles for my comic panels! How sharp can it get with detailed scenes?

Jaime

@cruise_chen it's pretty sharp. You should give it a try at labs.google/whisk