Stages AI previews STUDIOX for real-time V2V production
Dustin Hollywood says Stages AI V5 will stream iPhone or Fuji footage into real-time generators for V2V production. The preview includes mocap, depth-map rigging, Gaussian splats, and AR/VR worldbuilding.

TL;DR
- STUDIOX is framed as a live production system: dustinhollywood's V5 preview says iPhone and Fuji XT-5 feeds can stream into realtime generators for V2V, mocap, depth-map rigging, Gaussian splats, and AR/VR worldbuilding.
- The working demos are phone-to-screen tests, where a CUE workflow clip shows iPhone capture becoming a stylized live output and another test shows an avatar tracking motion and facial expression.
- The orchestration layer is the hinge: STAGES' official site says the product combines generation, editing, asset management, agents, and collaborative command workflows, while dustinhollywood's MCP update says MCP is in final security checks for V5.
- The rollout calendar has two public dates: a teaser puts something on 7/28, and the partnership post says a long-term partner will be announced on Aug. 3, 2026.
- Stages is already being used as a model-testing surface, with a Seedance versus Happy Horse comparison and a Qwen clarification sitting next to the STUDIOX previews.
STAGES' official about page describes the product as an operating system for cinematic AI production, with one command surface for agents, generation workflows, asset review, and delivery. The Model Context Protocol docs define MCP as an open standard for connecting AI apps to external tools and workflows, useful context for the repeated “tied to our MCP” language. ByteDance's Seedance 2.0 launch post says that model supports text, image, audio, and video inputs, while Hollywood's horse test said Happy Horse 1.1 looked more realistic in one Stages comparison.
STUDIOX realtime capture
Dustin Hollywood, whom the STAGES about page lists as Founder, CEO/CCO, put the preview in hardware terms: iPhone streaming now, Fuji XT-5 tested earlier.
The claimed STUDIOX pipeline has five pieces:
- Live camera feed from an iPhone or Fuji camera.
- Realtime generators for V2V production.
- Motion capture from the same live feed.
- Pre-viz conversion into model rigging from video depth maps and Gaussian splats.
- AR/VR worldbuilding, tied to STAGES MCP.
Hollywood said the system is coming in STAGES V5, with the broader STUDIOX release aimed for fall.
Phone-to-avatar tests
The strongest evidence is the loop itself: live body movement, a phone feed, and a generated character output running beside it.
The demos show three mechanics that matter for production:
- Capture: an iPhone records a dancer or performer.
- Transformation: CUE displays a stylized V2V output on a desktop monitor.
- Puppeting: a digital avatar follows body movement and facial expression in realtime.
For filmmakers, the useful jump is that capture, style transfer, and rig control are shown as one loop, not three separate handoffs.
MCP layer
Hollywood said STAGES MCP is going through final security checks and polish, with availability planned for the V5 update.
The official MCP introduction defines MCP as a way for AI applications to connect to data sources, tools, and workflows. In dustinhollywood's STUDIOX preview, the live capture, V2V, rigging, splats, and AR/VR pieces are described as “all tied to our MCP.”
That makes MCP the routing layer in the preview: the place where CUE can call production tools instead of asking a creator to hop between generation apps.
Model mix: Seedance, Happy Horse, Qwen
STUDIOX is arriving inside a product already built around model comparison and model routing.
Hollywood posted a horse-generation comparison and challenged viewers to tell which clip was Seedance and which was Happy Horse. In his follow-up, he said the more realistic shot was HH1.1, adding that Seedance is “amazing” but not best at everything.
A separate 7/28 teaser was clarified as Alibaba Qwen, not KREA, in the Qwen reply. The point is practical: STAGES is being shown as a command surface for multiple generators, not a one-model demo reel.
Look tests around the system
Hollywood also posted a run of Stages-made image and film tests around the STUDIOX preview window.
The examples range from sunlit fashion images to surreal editorial scenes and glitchy cinematic sequences. The STUDIOX pitch lands differently against that backdrop: live capture would feed into an aesthetic system already being used for polished look development.
Dates on the board
The near-term rollout is split across teasers, MCP, and a partner announcement.
Known dates and windows:
- 7/28: one teaser posts a date without naming the release.
- V5 “very soon”: the MCP update says MCP will be available in the V5 update after final security checks.
- Aug. 3, 2026: the partnership post says STAGES will announce a long-term partner.
- Fall: the STUDIOX preview says the bigger STUDIOX system is coming for STAGES users this fall.
Coffee-cup AR layer
AmirMushich sketched the non-film version of the same shift: live video becoming an interface for product information, quests, and rewards.
His example compares a 2019 coffee-shop AR app with 100-plus downloads to a 2026 workflow that can run lighter inside live video:
- Product information and brand stories overlaid on the camera feed.
- Real-time challenges and reward systems without a separate app.
- Personalized information, challenges, and rewards per viewer.
- Retention and conversion mechanics attached directly to the live visual layer.