Hey now, you're both pretty! Hermes is awesome. Claude is great too. Most people need easy to set up, out of the box solutions. Hermes isn't that. We use Hermes, ChatGPT and Claude here because different solutions need different tools.

RunDiffusion supports architecture practices of every size, from Large to Small and even Solo studios. When you’re managing design work, revisions and client presentations, having multiple AI tools in one place can make a practical difference. RunDiffusion brings image and video models together so you can explore concepts, compare materials, refine existing renders and create presentation visuals through one platform. For larger firms, Enterprise governance controls add user permissions, usage visibility and identity integration to support adoption across teams. In a personalized demo, we can discuss your projects, explore relevant design and visualization workflows, and review the options that fit your team size and governance requirements. Book a demo to see how RunDiffusion could support your architecture practice: rundiffusion.com/industries/…

Congrats! Can't wait for the team to try this out!

We would love to evaluate this for not only internal use but for use by our clients and customers. We are the leaders in providing inference for architecture firms. Our largest deploy is 4,000 architects in a single enterprise. Trying to win another deal from runway. Help us! :)

AI models are going to keep getting better. New capabilities are opening up more useful ways to explore materials, refine existing renders and create visuals for client presentations. For your firm, the opportunity is to put those improvements to work on real projects. RunDiffusion brings multiple image and video models into one platform, with Enterprise governance controls to help firms manage access, usage and budgets. Your team gains flexibility as the technology develops, supported by oversight at the firm level. In a personalized demo, we can discuss your firm’s priorities and explore how RunDiffusion could fit your design and visualization workflow, from early concept studies to image editing and architectural video. Book a demo to explore RunDiffusion for your architecture team: rundiffusion.com/industries/… #Architecture #ArchitecturalVisualization #RunDiffusion

A client asks to see another façade material before the next design review. Your team can use an existing render to explore that option on RunDiffusion. Start with the image, mark the area, and describe the revision as you would to another designer: “Edit this architectural render. Change the upper façade inside the red outline to vertical timber cladding. Keep the windows, ground-floor glazing, building shape, camera angle and daylight unchanged. Match the existing perspective and remove the red outline from the finished image.” For architecture firms evaluating AI, a familiar revision like this provides a concrete starting point: one image, one requested change, and a result your team can evaluate against a clear brief. Request a RunDiffusion demo to explore image editing for your architecture team. #ArchitecturalVisualization #Architecture #RunDiffusion

We had the pleasure of sitting down with Andrew Rider, Associate and Key Manager of Design Technology at BWBR Architects, to discuss how AI is supporting architectural design and client communication. One part of our conversation focused on a challenge many architects will recognize: seeing the potential in an early concept and helping a client see it too. A rough model or sketch can communicate a great deal to a trained designer. A client may need more visual context to understand the materials, atmosphere, and character of the proposed space. Andrew shared how AI visualization can help close that gap, allowing teams to present more developed visual studies while the design is still in its early stages. It’s a practical use for RunDiffusion: helping architects explore and communicate design directions so clients have something concrete to respond to, while leaving room for feedback and refinement. Thank you, Andrew, for sharing your experience and perspective with us. Explore the full conversation on RunDiffusion's Podcast: rundiffusion.com/podcast #Architecture #ArchitecturalVisualization #RunDiffusion

Does AI Take Away the Creative Part of Architecture? The creative part of architecture is exactly what many practitioners worry about losing to AI. That concern deserves a serious conversation, particularly when a firm’s design approach is central to its work. There’s value in exploring how AI fits into an established studio process: developing an idea further, investigating an alternative, or making a concept easier to discuss. The interest is in what it adds to the team’s own thinking. For architecture firms considering AI, creative authorship is part of the adoption decision. The tools need to fit the way the practice develops and evaluates design. RunDiffusion brings together AI image generation and editing tools for teams exploring that fit. Explore RunDiffusion for your practice at rundiffusion.com. #Architecture #ArchitecturalDesign #RunDiffusion

What could your next architectural concept look like in timber, glass, or a planted façade? Flux 3 Image is now on RunDiffusion, with native 2K and 4K image generation and editing for exploring those visual directions. This showcase moves through planted towers, curved interiors, and sweeping timber canopies. Put it to work on your own brief: explore a material palette, develop an exterior concept, or refine the atmosphere of an interior before choosing a direction to present. Try Flux 3 Image on RunDiffusion with a concept you’re working on. #Flux3 #RunDiffusion #ArchitecturalVisualization

Ideogram is legit! Always a favorite on our platform. Architects make a lot of edits and our customers are going to love using this! Congratulations @ideogram_ai

The same thing happened when Fable 5 was in the rotation and Opus 4.8 was the best opus model. Opus 5 was released and surpassed Fable 5. It’s a “tick tock” pattern they keep doing. Fable 5.1 still thinks differently and does things a little differently. Still good to use.

Such a fantastic offering for a team to have a cheap, mid tier, and max offering for getting work done. About 75% of our chats are “context gathering questions” to set up a big job to be handled by Astra / Sol. Great work from the OpenAI team!

Astra and Opus couldn't be farther apart in how they work. Clearly

Just wanted to recognize how legendary you are. Keep up the great work.

Restart your computer. Our devs have been saying that helps trigger the update.

Restart your computer. For some reason all our devs had to do that, then the updates started hitting.

Open Video Edit, select Gemini Omni Video Edit, and upload your clip. Describe the changes you want, then click Run. In this example, the edit changes the gallery’s interior walls and columns to natural oak with a warm light-brown finish and visible wood grain. The prompt also specifies what to keep: the original shapes and positions, ceiling, windows, artwork, lighting, and camera angle. When writing your own prompt, name the surfaces you want to change and describe the material, colour, and texture. Include the details you want preserved so the request is clear. Follow the walkthrough to try it with your own footage on RunDiffusion. #ArchitecturalVisualization #ArchViz #Architecture #AIVideoEditing #RunDiffusion

How do you get stuff out so fast?! Amazing 🤩

Don’t even release the weights at that point. Just keep it behind APIs. When we had to keep Juggernaut Flux behind a license (because of the parent Flux license) we just kept the weights private and launched with Fal/Runware. It worked great.

We would love that! We have a product where speed, cost, and accuracy is a delicate balance and a strong model can put it all to bed and we can ship it. Luna isn’t smart enough. Astra is too expensive. Deepseek v4.1 Flash seems to be the sweet spot but we’d take a slightly smarter model for a little more money if it worked better. Union Alpha was our hope. Should we DM you? Are these Openrouter credits? Or through your own inference endpoint?