This paper illustrates in detail the construction and deployment in the field of a low-cost video (and image) recording device for monitoring plant-pollinator interactions. The system is based on low-powered components that are widely available, overcoming the most important limitations of similar systems, namely size, weight, autonomy and cost. The authors demonstrate the potential of PICT in two field experiments, one with hard-to-access flowers and the other with rare visitors, and in both cases the system performed well. The authors also include a comprehensive practical guide for building and operating PICT, as well as sample videos.

MAIN ARTICLE
Line 25: no need to use $ or € when specifying the currency already (USD or EUR). Please check this throughout manuscript.
Line 36: "extensively" is too strong an adjective for a system that, however promising, has only been tested over a handful of weeks. Please tone down but maintain overall meaning.
Lines 56-59: most of the references are about orchid pollination, consider swapping one or two with studies on other groups or angiosperms in general.
Lines 86-102: it is unclear where the authors refer to infrared CT systems and to motion-triggered CT systems. Consider reordering the sentences or rephrasing slightly.
Lines 124-125: in FIgure 1's legend it is not clear that an epoxy glue and not self-adhesive Velcro strips were used, please clarify as this is not obvious.
Line 129: "standard ¼ screw" please specify unit measure (e.g. in).
Line 197: perhaps plural (e.g. conditions)?
Lines 207-217: it is not clear how many PICT units were used for observing ebony flowers, please specify.
Lines 235-239: please indicate specifically the pages in the guide included as Supplementary information concerning the MATLAB script. Without any benchmarking, the comparison of computer vision methods can only be qualitative, please make this clearer.
Line 242: it could be useful for the reader a brief sentence at the end of this paragraph summarizing in which cases automated motion detection can be useful (e.g. sporadic visits) and in which it is not (e.g. when visits are very frequent or very long).
Lines 278-279: in Figure 3's legend, although it could be obvious, consider adding that the bars refer to visitation frequency and the boxplots to visit duration.
Line 373: DOI link appears to be broken.

SUPPLEMENTARY MATERIAL
Pages 25-26: authors report that PICT by default creates 100 videos of 1 h each, and this can be modified by the user. This is important information that should be mentioned in the main article's text, as some use cases might require longer clips.
Pages 25-26: sections 2.9c ("Example script to continuously record video") and 2.9d ("Example script to record images at regular interval") seem to be swapped: the script provided in 2.9c records snapshots at 5s intervals, and the script in 2.9d records video clips. Also, consider adding brief comments to the code for video recordings (currently 2.9d), as less experienced readers would not know what the last two code blocks do.
Page 30: there appears to be some text missing from some boxes in Figure 22.
Page 36: some of the MATLAB code doesn't have a coloured background like the rest.
Pages 34-6: sections 3 and 4 of the guide explain how to apply two motion detection methods to automate visitor detection. Perhaps a brief introduction (a <5-10 lines total) to automated image processing and its potential pitfalls could be helpful to the readers to decide whether it is something they want to pursue. Furthermore, there are many computer vision methods available, and the field is developing very fast, so it could be worthwhile to encourage the readers to check the literature with some keyword combinations (e.g. "automated motion detection").
Some figures (e.g. Fig 10, Fig 14) have double bounding boxes or arrows/text, please check throughout.

SUPPLEMENTARY VIDEOS
The colour balance seems shifted toward reds in most clips, but not all, notably those in Video S4, that are also recorded @24fps. Was this an artefact due to hardware? While accurate colour reproduction may not be critical in many cases, it could be worthwhile to mention in the guide options available for colour correction, should this be necessary.