Joey breaks down a field test: shoot in a remote desert, upload via Starlink, and lean on Blackmagic Cloud and Eddie AI to turn footage into a finished cut in under 24 hours. He also cover handling long continuous clips with Claude Code, how Eddie AI organizes B-roll, and practical limits encountered when doing camera to cloud on the move.
The goal — speed and remote reliability
Joey’s core experiment was simple on paper and messy in practice. The objective: rapidly capture location footage around Borrego Springs (where we tried to track down every location in the film One Battle After Another), upload camera proxies from the field, and accelerate editing with AI-assisted tools so the whole project could be delivered fast. The test combined a Blackmagic PYXIS camera, a rented Starlink connection, Blackmagic Cloud sync, and Eddie AI for B-roll analysis and assembly.
Blackmagic PYXIS and the reality of on-location connectivity
Joey filmed on the Blackmagic PYXIS because of its native camera to cloud workflow. The camera lacks built-in Wi-Fi and a USB-C Wi-Fi adapter, so the team relied on the PYXIS Ethernet port to send proxies to Blackmagic Cloud.
Starlink provided internet in the middle of the desert. The setup included the Starlink dish feeding a modem and a battery pack that powered both Starlink and a laptop. An Ethernet cable ran from the modem into the PYXIS so the camera could upload proxies directly to Blackmagic Cloud. Audio from a Tentacle recorder and other media were pushed to the same cloud project so the editor could start working almost immediately.
How Blackmagic Cloud enabled distributed editing
Because proxies were uploaded to Blackmagic Cloud, the editor could begin work without waiting for full-resolution media. That split—camera capturing on location while the editor receives proxies in the studio—was the foundation for a fast turnaround. The test showed how cloud-first workflows let production overlap shooting and editing, but only if the upload speed holds up.
Eddie AI — speed comes from smart B-roll
For this project, Eddie AI was used mainly to analyze and tag B-roll, then assemble stringouts of the “best” moments. Eddie reads proxies inside a Blackmagic Cloud project, creates descriptive subclips, and generates assembled timelines of highlights that editors can pull into a cut quickly.
Key Eddie AI capabilities highlighted:
Automated subclips and descriptive labels — long takes are split into searchable segments with metadata-driven names.
Find-best-subclip stringouts — Eddie can export a condensed sequence of what it determines are usable, non-shaky, visually interesting moments. Great for montages and transition beats.
Text and object recognition inside frames — Eddie picked up signage and map labels (for example, Anza Borrego) and surfaced those clips when Joey needed them.
Handling long rolling takes: Claude Code and FFmpeg
A major pain point was continuous Insta360 dash footage: hours of rolling material that need to be scanned for usable beats. Eddie AI imposes a ten-minute per-clip limit for B-roll analysis, so Joey used Claude Code to run a script that split long takes into ten-minute chunks. The script relied on FFmpeg and a few helper packages to automate the chopping, then fed those subclips into Eddie.
This small automation step changed the workflow from manual slog to systematic pipeline: long continuous footage becomes many short, analyzable clips with metadata attached, which speeds both discovery and assembly.
What worked well and where the workflow hit limits
The combined toolset proved powerful for post-shoot rounds of refinement. Eddie excels when the job is to find specific shots during second, third, or fourth pass edits—especially montages, transitions, or atmospheric inserts.
There were two notable limitations:
Upload inconsistency with Starlink — first setup achieved upload speeds near 30 Mbps, but a later setup dropped to 3–4 Mbps and sometimes crawled to kilobytes per second. Multiple simultaneous uploads across devices compounded the issue. The team had to abandon the second upload attempt because it was too slow to be practical on location.
Camera hardware friction — the PYXIS requires Ethernet for cloud uploads. Built-in Wi-Fi or a native USB-C Wi-Fi adapter on the camera would remove setup friction and enable more nimble, on-the-move uploads.
Practical alternatives and mitigations
For longer-term mobile workflows, consider a travel Starlink mounted semi-permanently to a vehicle roof to maintain continuous upload throughout the day.
Limit concurrent devices on the local network during uploads and prioritize camera proxy transfers.
Use automated splitting (Claude Code + FFmpeg) for any source that records continuous long takes.
Closing Thoughts
The field test showed clear promise: Blackmagic Cloud plus Eddie AI can compress editing timelines when proxies and metadata are flowing reliably. The sticking points were connectivity and some hardware limitations. For teams moving toward always-on cloud workflows, the priority should be reliable uplink design and automations that adapt camera output to AI tool constraints.
Joey ends by asking producers to share similar workflows or questions, signaling that this is an iterative conversation rather than a finished blueprint. The practical lesson is straightforward: combine cloud sync, AI-assisted organization, and small automation scripts to get more editing done earlier in the production cycle, but expect to troubleshoot connectivity in the field.
