Announcing Roboflow Rapid: the first prompt-based model creation engine. Data labeling is dead. Go from idea to deployed model in minutes without labeling data. Upload a video, type a text prompt, and get an API. No data labeling teams. No manual annotation. No infra / dependency hell. If you... are spending 95% of your time labeling data and only 5% building your app, you are doing it wrong.show more

The barrier to building with visual AI just hit zero. And we are giving all users 2,000 free credits through the end of 2025 to try Rapid.

step by step tutorial on how to use rapid

So it uses SAM to automatically train a smaller/faster model (like Yolo or RF-Detr)?

yes! rf-detr for the detector key thing is we can always use the best foundation model and detector for the task. adding in more models over time (ie Gemini Pro 3, etc) will ensure this continues to be the best possible path for people to make vision models fast

Thanks! This is such a neat project and idea. This lowers the barrier to entry for people to dip their feet in computer vision and create something that could be used in a factory, for instance. I work at a university, and I'm going to share this with both the manufacturing engineering department and the data science department.

yeah that's exactly right if there is an HMI attached to a camera on a line, you could make a model quickly at the edge for a new issue/product very easily

No more labeling hell? Bold move. Edge cases are the real test though.

if it actually works, removing labeling would be a huge productivity boost for builders

try it out and let us know

@vikhyatk 👀

Congrats team! 🙌

Impresionante!!! Gracias por este tremendo regalo de navidad 😊

any projects in mind to try it out with?

Love this and thank you. However, I find Roboflow Rapid fails on complex annotation tasks with subtle differentiation. It’s best for clearly visible objects rather than, for example, different classes of road damage

great example of an application that would require fine-tuning a custom detector, totally agree

Your reply is validating - thank you!

rapid can probably handle 80%+ of objects people want to detect for vision applications, specific ontology like this is in that 20% :) Looks like a cool project you're building. Tons of valuable use cases in understanding local infrastructure

Just wondering if it can be used in cell type detection and counting for blood test labs?

that will likely work. give it a try, it's totally free, takes just a few minutes

Thnx, will do

lmk how it goes

How to subscribe to growth plan for dedicated online support? Please reply to my email id: [email protected]

Prompt-based model creation is intriguing. Curious how well it handles edge cases and complex tasks. Anyone testing those limits?

very data dependent best way to find out is to try it and share results, it's free what use cases do you have in mind?

How to subscribe to growth plan for dedicated online support ? My email id: [email protected]

Hello! Please send any support questions to Email support is only available for Enterprise plans

Then how to subscribe to enterprise plan ?