A 2.5" drone frame. Star shaped dual arm design available both in printable and carbon fiber cut files in [[STL]] and [[CAD]] formats.

This project was born out of the desire for building a robust freestyle 2.5" 2S quad built from the ground up around the HDZero AIO15 and Lux camera. Something with a bit more power and wind resilience than a whoop, but still relatively quiet and survivable to crashes. Some of the mono-body frames seem to have lightweight arms that are a bit flimsy, thus the star shape was opted in the newer versions of the Kumo frame.

My vision for this project was to create an open-source frame design that could be iterated and shared with accessible options.

1. Elements
2. Design
3. Recommended Setup
4. Tune
5. Videos
6. Authors
7. Versions
8. Changelog

• Open Source as CC-BY 4.0
• Designed as a more robust toothpick with AIO or other light stacks. Primarily the HDZero AIO15
• Squished X geometry (4mm stubbier to get the props a bit less in frame)
• Slam deck to keep center of gravity along propeller line
• Rigid Design
• Camera mount designed for 25/35deg HDZero Lux. Also compatible with the Kayoumini camera mounts

Cutout for HDZero AIO15 to reduce height while also keeping ports accessible.

Squished X geometry set at just under 85deg between front/back. Reduces prop in frame, however roll vs pitch movement requires less inertia due to weight distribution along Y axis.

Keeping the center of gravity low with 8-12mm standoffs helps reduce effort in flips. Sweet spot for avoiding battery strikes with props sits around 10-11mm standoff while also providing plenty of room for wire routing. For a tighter fit, 8-9mm standoffs can be used. Photo below is of 11mm standoff.

Designed with compatibility in mind with camera mounts from Kayoumini This sample mount for HDZero Lux at 35 degrees offers a balance between racing and freestyle. Print in TPU to dampen vibrations and soften impacts.

Original test builds used a variety of 3D printed versions yielding varying stiffness and weight. Thickened printed version in PETG-CF to increase stiffness and improve impact resilience. In this version I was still using the stick antenna from BetaFPV

Stubby version of frame cut from 2.5mm Carbon Fiber with RHCP antenna for improved weight and video signal

Noise comparison between tunes of 3D printed (blue) and Carbon Fiber (red) variants of frame. The CF frame seemed to produce close to 10dB less noise thus requiring much less filtering to improve responsiveness.

I created a light weight version of the frame for increased agility (and likely more fragile) stubby version weighing in at 10g for the base. Cut from 2.5mm thick CF 6K 2x2 Twill weave

And a heavier 3.4mm carbon (was listed as 3.1) cut from 6K 2x2 Twill weave. Added protrusion to add material in front of motor. Designed for more aggressive bashing, but adds 4.4g to previous design.

TPU plus CF stubby, top plate, standoffs and main screws comes in at 17.5g

Top plate cut from 1.5mm CF 6k 2x2 twill. Designed with cut out towards the front for battery lead and extension on the back for battery deck/attachment for antenna.

Key considerations for building:

1. Weight of components: Antenna type, TPU mounts and motor guards, battery size. My goa was to have a dry weight under 70g and an AUW of under 100g. For this I found the stubby frame at 2.5mm thick with a circular polarized antenna provided a decent compromise in weight, range, signal quality. For a more racing style setup, I tested the Twiglet frame that allowed me to get as low as 50g using 12025 11000kv motors
2. Power: For motor and battery size considerations, I found the HDZero 15A felt solid with 2S in terms of maneuverability. 3S seemed to add quite a bit of weight which required much more throttle to power through sharp corners. With the 2S battery, finding a balance between 8000-11000kv would likely be ideal. I opted for heavier 1104 7500kv which added 4g, but gave me a bit more control with 2" pitch props over 12025 11000kv. That being said, the 12025 felt faster and twitchier, which may be preferable for pilots with better reaction time.
3. Props: I tested with GF 2520, 2540 and AVAN Rush (1.9" pitch). By performance, it felt like the Gemfan 2520 yielded the best performance and was less noisy than the AVAN rush. As nice as the AVAN rush felt, it had a high pitched scream that was somewhat unpleasant (and may be a nuisance around others). The GF2540 was explosive in power and worked well with the 1104 motors, but seemed better suited for racing and higher speed lines.
4. Antenna: The default stick antenna that comes with the HDZero AIO15 caused a bunch of problems and no matter where I placed the antenna, I always seemed to struggle with noise and range issues. Opting for a circular polarized antenna that I could push further away from the body seemed to help quite a bit in transmission quality during punch-outs. I had tried other stick antennas such as a BetaFPV linear antenna for Air75, but performance still seemed quite bad.
5. Accessories: The HDZero AIO15 does not come with a blackbox, which means if you wish to tune, you will need to attach an external recorder. I used a Flywoo OpenLager, which helped narrow down on some of the noise and responsiveness to cut down on propwash. I also added a simple buzzer as this tiny thing is super easy to lose. Adding 1g worth of wire/buzzer/ziptie is totally worth it in my book.

• HDZero AIO15 + Lux camera
• T-motor 1104 7500kv
• Gemfan Hurricane 2520
• Newbeedrone RHCP Honey Dipper Antenna
• 2S battery (450-580mAh)
• Active Buzzer and Flywoo OpenLager blackbox

Betaflight defaults were fairly decent for me, but minor tweaks have helped improve propwash. Carbon fiber frame also was less susceptible to vibrations and required less filters.

PID profile settings are almost stock. I used PIDToolbox and suggestions from this video. Angle limit was set to 30 during auto-wobble script.

For the filter, removing the first lowpass and tightening the notch filters helped reduce delay while still cleaning the noise.

I also set the filter weights and q value to the following: rpm_filter_weights = 100,60,70 rpm_filter_q = 1000

Resulting first harmonics were still present, with Roll requiring a bit more attention, yet fairly solid from what I can tell.

Kumo Frame analysis for stress concentration: (front impact)

Kumo Frame printed first flight

Kumo Frame CNC cut on Millennium Mill

The Kumo project was designed and contributed by:

• Kumoi: primary designer
• Auburn: Slim 3D print version & design contributions to main

For a list of past created iterations, here is a Google sheet of various versions with specs, components and photos List of file versions in repo:

• Kumo Main Version (Sharp version)
• Kumo Slim (less material, more fragile)
• Kumo Stubby (similar to sharp version with stubby motor mount)
• Outdated versions (single arm design):
  • Kumo Frog (quite heavy and poorly designed skeletonized parts)
  • Kumo ChunkyFrog (more material, more cutouts, poor access to usb port)
  • Kumo Offaxis (interesting attempt to merge top/bottom plate with flipped parts. Too heavy, poor skeletonization, no top/bottom usb access, no battery strap)

• 6-19-2026: Additions of new slim 3D printable versions by Auburn
• 5-21-2026: Added Git Discussion, fixed M2 clearance on bottom AIO hole (Thank you Auburn!)
• 5-19-2026: Completed Documentation V1 and published to Reddit
• 5-5-2026: CNC cut stubby frame (youtube video above)
• 4-24-2026: Tested with 3S... a bit scary
• 4-20-2026: First flight of Star shaped Kumo frame (youtube video above)
• 4-19-2026: 3D printed version of frame built
• 4-16-2026: Experimental version of 3D printed frame. Felt very flimsy, added some material around standoff mounts.
• 4-6-2026: Camera mount complete
• 3-28-2026: Experimenting with Flipped Z design
• 3-27-2026: Stress tests for impact yields concerns with single arm design
• 3-25-2026: First version of Kumo frame designed (looks like a squished frog)