Red River Insights - Visual Production Stack top 10
From product photos to digital twins, AI is turning visual work into infrastructure.
Red River Insights - June 2026
Dear friends,
By visual production, we mean the workflows used to create, capture, edit and deploy images, videos and 3D assets across commerce, architecture, product design, gaming and simulation. This used to be a craft bottleneck: a skilled person had to shoot, edit, render, model, texture, optimize, convert, review and export each asset through tools that were powerful, but rarely connected.
That bottleneck is starting to break. The first-order AI story is obvious: models can generate better images, videos and 3D assets. The more interesting story is what comes after generation: the workflows that turn visual work into repeatable infrastructure. When this works, the payoff is concrete: faster turnaround, lower production cost, and assets that can be reused across formats and channels instead of rebuilt from scratch each time.
That is why this is not a pure “3D imaging” ranking, and why the scope deliberately spans several industries: e-commerce, architecture, interior and product design, hardware engineering, gaming, simulation and physical AI. Their needs are not identical, a fashion catalogue and a video game ask very different things of the same pipeline in terms of precision, realism, file size and approval workflows, but they share one software layer. Visual Production Stack is the layer that captures real objects, generates or edits visual assets, preserves product fidelity, lets teams collaborate around spatial files, and moves outputs into downstream systems such as CAD, BIM, e-commerce, AR, game engines and simulation environments.
The strategic question is whether European workflow companies can lock in the professional layer (the industry-specific formats, integrations, fidelity and collaboration that real workflows depend on) before the model layer bundles more of the process. Runway, Midjourney or Black Forest Labs are all moving from raw generation into workflows, APIs, custom models and tooling. The defensible companies in this market will need to own the messy layer around the model: fidelity, formats, integrations, QA, versioning, collaboration and the source of truth for real-world assets.
RAMP’s Visual Production Stack top 10
We hope this sparks interesting conversations. If you have comments or would like to suggest a startup that should be included, feel free to reach out. Joseph, Chloé and Olivier will be delighted to discuss these trends and rankings.
(Ranking established on 23/06/2026.)
We have highlighted four key trends illustrated by these companies:
1. Visual assets are becoming production systems.
The clearest shift is from one-off creation to repeatable visual pipelines. Photoroom now positions itself as a full AI visual solution for e-commerce, with batch editing, Image API, marketplace sync, product staging, brand kits and custom models for catalogue-scale teams.
Allsides and Solaya show the same pattern in 3D. Allsides says its scanners can mass-produce reality-grade 3D digital twins and generate more than 30,000 models per scanner annually. Solaya starts from a two-minute smartphone scan and turns one product asset into images, videos, 360 views and AR experiences.
2. Spatial design is having its collaborative software moment.
Architecture, interiors and hardware engineering are moving from specialist files to shared visual workspaces. Gendo turns sketches and design inputs into photorealistic architectural renders inside a collaborative canvas. Rayon combines CAD, BIM-like drawing tools, AI visualization, AI tracing and realtime collaboration for interior designers and architects. Encube Technologies applies the same logic to hardware engineering, bringing 3D models, drawings, measurements, feedback and permissions into one shared browser workspace.
In June 2026 Adobe added native 3D asset review to Frame.io, letting brand managers, agencies and marketing teams comment on and approve USD, GLB, FBX and CAD files without opening a 3D tool. That is a useful signal of where both the value and the competitive pressure are heading: the file becomes a shared decision surface rather than a specialist artifact. For the startups here, defensibility cannot stop at review and approval. They need to own domain-specific workflows, precision, integrations and the operational data around the asset. Adobe can connect creation through Substance 3D with collaboration through Frame.io, so the bar is not simply better collaboration, but deeper vertical workflow ownership.
3. The hard problem is not generation, it is trust and delivery.
Many professional workflows need accurate geometry, material fidelity, version control, fast loading and downstream compatibility. NavLive uses LiDAR and AI processing to scan building sites, then produce 2D floor plans, point clouds and 3D models for AEC teams. Its public examples show building scans completed in minutes, from a 2,311 sq ft residential property in 9 minutes to a 97,261 sq ft conference floor in 26 minutes. Capture hardware is not always required: lightweight cases can start from a smartphone scan, as does, or be generated from scratch, while precision-critical work such as AEC site surveys still calls for LiDAR. The right entry point depends on how much real-world accuracy the downstream use actually demands.
DGG / RapidPipeline sits downstream of 3D creation, providing the automation layer that turns raw 3D and CAD files into optimized, interoperable assets ready for real-time, web, XR and enterprise workflows. DGG says RapidPipeline has processed more than 6m 3D models and contributes to open 3D standards such as glTF, USD and MaterialX. These open standards predate today’s AI tools, but they matter even more in an AI workflow: they are what lets a single asset move cleanly between capture, generation, optimization and downstream systems without breaking. Threedium is pushing in the same direction from the creation side, with text or image to production-ready 3D, optimization and deployment across web, commerce, AR, games, simulation and rapid prototyping.
4. Model bundling will squeeze weak workflow layers.
The same technical problems appear across games, e-commerce, architecture, hardware and simulation: create the asset, preserve its fidelity, optimize it, collaborate around it, then deploy it into another system. Atlas targets games, CGI and visual worlds with an all-in-one AI workflow, including moodboard-to-3D and production-ready assets. Threedium bridges commerce, AR, games, simulation and web. DGG / RapidPipeline turns CAD and 3D files into marketing-ready and real-time assets.
If model providers bundle generation, custom training, workflow builders, APIs, asset libraries and collaboration, thin wrappers will be vulnerable. The European companies that last will need to own something deeper: domain-specific formats and integrations, product fidelity, CAD and BIM workflows, 3D optimization, QA, asset versioning, privacy, provenance or the day-to-day team workflow.
Open questions
Will visual production become horizontal infrastructure, or stay vertical by industry? Architecture, gaming, retail, robotics and hardware engineering all need visual and 3D assets, but their requirements are different. Precision, file size, realism, collaboration, privacy and approval workflows may pull winners toward vertical depth.
Who owns the source of truth for a 3D asset? An object can start as a scan, become a generated model, be optimized for web, edited in CAD and deployed into AR or a game engine. Ownership, versioning and provenance become strategic once the asset moves across that many systems.
Can AI-generated 3D meet professional quality requirements? For a hero render or a marketing image, increasingly yes, and the consumer tools are genuinely impressive. For production, not yet by default: AI generators still tend to output messy topology, guess at the geometry they cannot see, and miss the dimensional precision that CAD and engineering demand. There is also a fidelity gap that speed alone cannot close. As Adobe noted when it added 3D review to Frame.io, generative AI cannot guarantee that the handbag in the render is your handbag, with your stitch count and hardware finish; only a physically accurate digital twin can. Production teams need geometry that behaves, materials that render correctly, brand-exact accuracy, files that load fast and outputs that survive downstream tools. That distance between impressive and dependable is exactly where the defensible companies live.
Can workflow owners defend against model-provider bundling? Runway, Midjourney, Black Forest Labs and platforms like Luma can keep moving downstream into tooling. The open question is whether startups in this ranking can own the professional layer: integrations, QA, file formats, collaboration, versioning and physical sources of truth.
Reach out if you have an opinion or to suggest a company we might have missed!
More on RAMP’s scoring method
The ranking of these startups is based on estimated company momentum, but the algorithm does not assess the quality or reliability of the products and solutions developed by these companies.
RAMP’s Growth Score identifies the companies showing the strongest momentum. Red River West then applies a theme-fit and company-status check.
Find out about the algorithm behind this ranking and the way scores are calculated here: Cheat sheet on RAMP.
In case you missed them, our latest top 10s are here: Ocean Tech (May 2026), Agentic Commerce (April 2026), AI’s Picks and Shovels (March 2026), AI Safety & Governance (February 2026), and Energy (January 2026).
All the previous Top 10 are here.
Portfolio news:
The Exploration Company completed a successful drop test of its Nyx capsule recovery system, releasing a test vehicle from 2,800m over the Mojave Desert and validating the handover from drogue to four main parachutes for a nominal touchdown, and unveiled Storm, a new high-thrust reusable European rocket engine.
Ada published a peer-reviewed Nature Health study on its clinical AI inside South Africa’s MomConnect maternal-health platform, finding that more than half of the pregnant women and mothers who received treatment had not planned to seek care before using Ada, with safe health-seeking behaviour up by nearly a third.
Veesion expanded its retail-theft-detection AI, presenting at the Salon National Carrefour on 10-11 June and pushing into new segments and markets, including pharmacies and Latin America.
Okeiro was named to Galen Growth’s HealthTech 250 2026, the annual signal-driven selection of the most promising early-stage digital-health ventures worldwide.
HyPrSpace announced a partnership with Tumbleweed to expand access to rapid microgravity testing via its Baguette One suborbital launcher, with a first technology demonstration flight planned for Tumbleweed’s next-generation Pod payload container.
Sans Strings Studio signed with Creative Artists Agency (CAA) for representation across gaming, film, TV and licensing.
Otera was named an Enterprise Innovator in HFS Research’s Horizons: Agentic Technology 2026, with the company highlighting production deployments in regulated workflows.
Team news:
A lot happened in June for the Red River West team, let’s unpack!
First, we are delighted to welcome Estelle Godard as Principal in our investment team in Paris. Estelle joins Red River West after investing in deeptech and space at Promus Ventures and OTB Ventures, with earlier experience at the European Space Agency. We’re super excited to have her on the team!
True to our positioning as a Data Driven VC, Joseph co-hosted with OpenAI a hands-on workshop on how Codex can be used in VC. Thank you to the many VC friends who joined us for a very concrete evening of workflows, demos and shared notes.
A few weeks earlier, Chloé had also organized a workshop with the SheEOs association for female students and young professionals, showcasing how AI is reshaping our day-to-day work.
At the beginning of June, Olivier and Chloé were in Madrid for South Summit. Olivier joined a panel on digital sovereignty, while Chloé served on the startup competition jury and had the chance to meet a surprise guest, his Majesty the King of Spain!
Olivier also shared his perspective at Crédit Agricole CIB on deeptech investment and the role of European technology companies in shaping the future of industries, sovereignty, and long-term value creation; and just last week, he participated in a panel hosted by BNPParibas on the competitiveness of the tech European landscape.
The >commit team also organized an event with Zed Industries at Neon Noir around open source, AI, and the Paris builder community, bringing together founders and engineers for talks, demos, Q&A, and informal exchanges.
Finally, we are looking forward to the RAISE Summit in early July, where Red River West is a partner, notably through >commit on the summit’s open source track. If you want to know more or if you’ll be around don’t hesitate to ping us.
You received this newsletter because we met or worked together and we thought you might be interested in an update. If we got it wrong, feel free to unsubscribe with the link at the end.
Until next month,
Chloé, Olivier and the Red River West team