How insurers, law firms and forensic engineers use drone flights today, what the rules of evidence ask of the imagery, and how Omniris gets the flight flown, measures what it shows and proves later that nothing was changed.
Drone imagery has become ordinary in property claims and in the investigations behind lawsuits. Carriers build drone programs for catastrophe assessment and roof investigations; forensic engineers map the debris field of an explosion or a structural failure before the evidence is disturbed; a drone gives a faster first look at a masonry façade that engineers would otherwise inspect by rappelling down the building [16]. A trade report in October 2026 described drone inspections as having moved "from a novelty into standard practice" in claims [17].
What the imagery has to survive is the same in every case: someone will ask whether it is what it claims to be, whether it was changed, whether everything relevant was kept, and whether the measurements taken from it can be trusted. The Federal Rules of Evidence answer the first question with authentication (Rule 901) and, for copied data, self-authentication by hash value (Rule 902(14)) [1][2]. The duty to preserve is enforced by Federal Rule of Civil Procedure 37(e) [8]. Measurements and models that support an expert's opinion are tested under Rule 702 and Daubert [3][4].
Omniris is a drone operations marketplace, a processing stack and an evidence vault in one product. For claims and litigation work, five things matter most:
Post a job with the area, the shots you need and the dates; approved pilots near the site are notified, apply or are offered the job, and deliver into your vault.
Maps, textured 3D models, point clouds and 360° panoramas, with a ruler on the 3D model that is withheld when the model has no real-world scale.
Every file gets a SHA-256 fingerprint and a sealed custody history, timestamped by an independent authority, with signed PDF certificates anyone can check online or offline.
A legal hold freezes a file, folder, job, site, account or organization, including the flight log and airspace approval, with a locked copy in write-once storage.
Frozen snapshots, password-protected links with expiry, and a custody record of each download, so you can show what the other side received and when.
Omniris supplies the records, the measurements and the means to check them; the legal judgement, the expert's opinion and any certification stay with the people who make them. What the product claims, and what it doesn't sets out those boundaries.
Three groups fly, or commission, most of the drone work that ends up in a claim file or a court record: insurance carriers and the adjusting firms that work for them; forensic engineering firms retained by carriers, owners or counsel; and, less often, law firms commissioning documentation directly.
Routine commercial drone flights in the United States, claim flights included, are flown under Part 107, the FAA's small-drone rule [25]. State Farm received an FAA waiver in 2018 to fly beyond visual line of sight and over people to assess Hurricane Florence damage across four states [19], and the FAA still lists "damage assessments supporting disaster recovery related insurance claims" among the operations it considers for expedited emergency approval [18]. The insurance regulators' association notes the value of drones in reaching inaccessible areas after a disaster [20].
The October 2026 trade report gives the reason claims organizations keep investing: a claim file with "fewer touches", at a time when Sedgwick expects nearly a quarter of claims adjusters to retire by the end of 2027, and counted 23 US weather disasters of more than $1 billion each in 2025 [17]. It also reports a policyholder's objection that an aerial pass could not see damage to the underlayers and decking beneath his shingles [17]. A forensic engineer makes a related point: a drone "cannot probe a suspected hail impact site for bruising" [16].
Forensic engineers fly where the scene is large, unsafe or about to change: debris fields after explosions and structural failures, mapped without disturbing the evidence; damage in high-rise buildings; masonry façades otherwise inspected by rope access; windstorm and tornado damage; fire scenes; roofs; and, as further opportunities, accident reconstruction and pre-loss documentation [16]. The same article reports photogrammetry accuracy of about ±1.2 in (±3 cm) horizontally and ±2.0 in (±5 cm) vertically, noting that significant placement of ground reference targets is required for higher resolutions [16]. A National Institute of Justice study found that drones used alongside survey instruments cut crash-scene clearing time by 35 to 45 minutes and slightly reduced some measurement errors [12].
Practice guidance shapes the deliverable. The Scientific Working Group on Digital Evidence (SWGDE) asks forensic photogrammetry to work from the original or a bit-for-bit copy, document the method and report the uncertainty of every conclusion [10], and has separate guidance on drone forensics [11]. ASTM E1188 covers collecting and preserving information and physical items for litigation [13]; ASTM E860 covers examining items when testing could change them [14]; ISO/IEC 27037 covers handling digital evidence [15].
| Mission | Who flies it | When | What the buyer expects |
|---|---|---|---|
| Catastrophe response | Carrier drone teams, adjusting firms, contracted pilots | Days after the event, often under emergency approvals [18][19] | High-resolution imagery of affected properties to judge the severity of damage [19] |
| Roof and exterior claim inspection | Adjusters licensed as remote pilots, contracted pilots | Per claim, within the state's inspection window (Florida: any physical inspection within 30 days after the insurer receives the proof-of-loss statements) [23] | Every slope and elevation, measurements for the estimate, no re-inspection [17] |
| Re-inspection, appraisal, dispute | Carrier, public adjuster, appraiser, engineer | When scope or cause is disputed | One record both sides accept; before and after |
| Special investigation (fraud) | Special investigation units | Case by case | What existed, and when |
| Forensic scene documentation | Forensic engineers, reconstructionists | Before the scene changes [13][16] | Maps and 3D models with stated accuracy [16] |
| Construction defect | Forensic engineers, envelope consultants, experts | At retention, around destructive testing [14], on repeat visits | Façade and roof documentation, defect locations |
| Exhibits and demonstratives | Experts, litigation graphics teams | Before deposition or trial | Views a jury can follow, tied to the data [7] |
Part 107 requires a remote pilot certificate and keeps routine flights within visual line of sight [25]. State laws add rules on capturing images. Texas makes drone image capture lawful in a list of specific cases, including an insurance company employee capturing images of property in connection with underwriting or adjusting a claim, when the operator is FAA-authorized for that airspace; its entries for licensed engineers and surveyors require that no individual be identifiable [21]. Florida's drone surveillance statute has no insurance exception but exempts FAA-compliant aerial mapping [22].
Claim-handling rules reach the imagery too. Florida allows an insurer to investigate by "electronic photographs or video recordings of the loss", naming photographs or video taken with a drone among the accepted methods, requires a copy of any detailed estimate to go to the policyholder within 7 days, and requires a log of adjusters and claim communications [23]. California requires claim files detailed enough that events and their dates can be reconstructed [24].
A summary of the federal rules a drone image or model meets on its way into a case. State rules differ, and none of this is legal advice; it explains why the records Omniris keeps are shaped the way they are.
| Rule | What it asks | What drone evidence needs |
|---|---|---|
| FRE 901 [1] | Evidence "sufficient to support a finding that the item is what the proponent claims it is", for example by a witness with knowledge or by "evidence describing a process or system and showing that it produces an accurate result" | Who flew, when and where; how the image became a map or model; proof it is the same file |
| FRE 902(13), 902(14) [2] | Self-authentication of machine-generated records and of copied data by a qualified person's certification. The 2017 note: copies are "ordinarily authenticated" by hash value, and identical hashes "reliably attest" to exact duplicates | A hash recorded at collection, and a way to recompute it on the produced copy |
| FRE 702, Daubert [3][4] | Since December 2023 the proponent must show it is "more likely than not" that expert testimony rests on sufficient data and reliable methods, reliably applied | Documented method, inputs and measured error |
| FRE 107 [7] | Since December 2024: illustrative aids (photos, video depictions, computer simulations) may help the jury, are not evidence, and, when practicable, must be entered into the record | Views and models tied back to the underlying evidence |
| FRCP 37(e) [8] | Electronic information that should have been preserved, lost because a party "failed to take reasonable steps to preserve it" and that cannot be restored or replaced, can draw measures to cure the prejudice to another party, and the severe ones (an adverse presumption or instruction, dismissal or default) only on a finding of intent to deprive. The duty arises when litigation is "reasonably foreseeable" | A way to stop deletion the day a dispute is foreseeable, including files added later |
In June 2025 the federal Standing Committee approved a proposed new Rule 707 for publication; it was published that August, and comments closed on February 16, 2026 [5]. It would have applied Rule 702's reliability standards to machine-generated evidence offered without an expert. In its report of May 17, 2026, the Advisory Committee on Evidence Rules said it "does not recommend action on the proposed Rule 707 at this time." It revised the draft to cover "artificial intelligence" rather than all "machine-generated" evidence, added a notice provision, agreed the revision would need republishing, and chose to have it vetted by technology experts at its fall meeting [5]. That meeting was set for October 15, 2026, with panels of experts on both AI evidence and deepfakes; the committee's agenda for it records that the published rule was withdrawn and a modified proposal kept for further study [6]. Rule 707 has not been adopted and has no effective date.
On deepfakes, the committee concluded that "at least for now, an amendment to Rule 901 to address deepfakes is not warranted", while drafting a Rule 901(c) to hold in reserve: if an opponent presents evidence sufficient to support a finding that an item was fabricated by generative AI, the item would be admissible only if the proponent shows it is "more likely than not authentic" [5].
Under a rule of that shape, a challenged photo would need more than a witness saying it looks right. A fingerprint recorded at upload, sealed and timestamped by an independent authority on the day of capture, answers "when did this file exist, and has it changed since?" It does not prove what the camera saw was real, and Omniris does not claim it does.
Each use from the industry survey above, and what Omniris delivers for it.
| Use case | What Omniris delivers |
|---|---|
| Getting a pilot to a loss | On the Omniris cloud and the data center edition: post a job with the area, pins for the shots you need and the dates; nearby pilots are emailed; offer it to a pilot who has worked for you or pick an applicant; pilots' Part 107 certificates are checked against the FAA registry. |
| Catastrophe response | A job for each property, each reaching the pilots near its own site, and recurring jobs for properties that need repeat visits. Emergency waivers are the operator's to obtain. |
| Roof and exterior documentation | Photos with capture details, a roof map, a 3D model measured with the 3D ruler (face area, perimeter, tilt), thermal flights for possible wet insulation. |
| Recording damage, and where | Pixel-pinned annotations, transcribed voice memos, findings with evidence crops and a signed history. |
| Before and after | Fade, curtain and blink between two dates of the same site, or compute the elevation change with its uncertainty; an older survey you hold can join the site's timeline as dated imagery. |
| Forensic scene documentation | Maps, measurable 3D models, photo-real splats, point clouds including LiDAR, 360° panoramas. |
| Defensible measurements | Ground control and check points with measured RMSE; a ruler that refuses a model built without scale. The opinion stays the expert's. |
| Construction defect over time | Every flight of a site by date, comparison between flights and against a design model, findings tracked to repair. |
| Custody and authentication | SHA-256 fingerprints, sealed custody history, independent timestamps, signed PDF certificates, a public verify page, an offline checker. |
| Preservation and retention | Legal holds from one file to a whole organization with locked write-once copies; retention policies with signed disposition certificates. |
| Production | Frozen snapshots, protected links, evidence copies, certificates, a custody-event spreadsheet. |
| Exhibits | A public 3D link, still images from the 3D view, and data in open formats for the expert's own tools. |
Everything starts with the upload. Drone photos, videos, LiDAR data and documents go into the vault, Omniris's private file store, from the browser, the desktop app or the pilot's phone. Each photo's capture details are read and kept with it: date and time, position, altitude, camera and lens, and the drone's own records where the camera writes them. The moment a file arrives its SHA-256 fingerprint is calculated and its custody history begins (see Custody, authentication and preservation).
From what the pilot uploads, Omniris builds the products a claim or a case usually needs: an orthomosaic map with surface and terrain models, a true-scale overhead view of a roof, a lot or a debris field (Stitching paper); a textured 3D model, a measurable surface painted with the photos, and a Gaussian splat, a photo-real record of how the scene looked, for seeing rather than measuring (Digital twins paper); point clouds, including LiDAR (LiDAR paper); 360° panoramas for context; and thermal maps of surface temperature, whose roof-moisture wording cites the conditions and caveats of ASTM C1153 (Thermal imaging paper).
Before a model or map is charged for, Omniris checks the photos: how many, whether they carry positions, whether they form one connected flight, and whether the camera angles suit the product (and, for a Gaussian splat, whether any are blurred). A flight shot at an angle cannot become a map, and Omniris says so and offers the 3D model instead. These checks catch a bad capture while the pilot can still fly again, which matters when the scene is about to be cleared or repaired.
When you post a job you can drop numbered pins on the job map, each with a note describing a ground-level photo you need ("north elevation, gutter damage", "close-up of the failed connection"). The assigned pilot sees them on the map and in a numbered list, on the web and on the phone. It is a simple way to turn an adjuster's or an engineer's checklist into the pilot's shot list.
Omniris can run open-vocabulary object detection over photos and maps: you describe what to look for in words, and the detector marks what it finds. A person reviews the result, corrects it, and promotes what matters to findings; every check or correction is recorded with who made it and when. Detections are a starting point for review, not a damage determination. See the AI detection paper.
A measurement offered as evidence has to come with its method and its error. Omniris is built so that a number is shown only when the data can support it, and the error is stated where it can be measured.
Open a textured 3D model and turn on the ruler. In distance mode, click points on the model: the readout gives the last segment, the total path, the straight-line distance and, separately, the vertical difference, so a "wall height" is never a diagonal. In area mode, click around a face (a roof slope, a wall, an opening) for its area, perimeter and tilt from horizontal; if the points are not on one plane, it says the area is a projection that under-reads a folded surface.
The ruler is withheld when the model has no real-world scale. A model's scale and place on the Earth come from the GPS positions the drone recorded: the record of how Omniris built the model shows whether the reconstruction was tied to them, and for Omniris's own 3D models how well the reconstructed camera positions agree with them. A model built without usable GPS, or whose placement failed, gets no ruler, and the viewer says why rather than measuring in an arbitrary unit. A model uploaded from elsewhere, with no build record, can still be measured but is not placed on the map. Gaussian splats are never measured; a point cloud on its own is measured only when it is in metres.
Illustrative example values. Readouts follow each person's unit preference: feet and square feet by default, metres and square metres for those who choose metric.
On a map, the measuring tools give distance (horizontal, and along the slope and terrain where elevation data covers the line), area and perimeter, an elevation profile along any line (exportable as a spreadsheet), and volume above or below a base you choose. Saved volume measurements carry their own custody events each time they are created, edited (recording the old and new values) or re-evaluated, and every volume is shown with its uncertainty. See the Site management paper.
Ordinary drone GPS is far less precise than a survey. A map can be tied to surveyed ground control points, and its accuracy measured at independent check points: surveyed points left out of the fit, so the fit cannot pull them toward their surveyed positions. The results give each check point's error east, north and up, and the map's horizontal, vertical and 3D root-mean-square error (RMSE). Check points added after delivery are measured against the finished map's saved reconstruction, and every measurement is kept with who asked for it. That is the documented error SWGDE asks examiners to report [10] and Rule 702 asks proponents to show [3].
Omniris computes and reports what the data supports, with its measured error. It does not stamp a survey, and a drone map is not a boundary or licensed land survey. Where a matter needs one, a licensed surveyor or engineer makes that call, and can use Omniris's check-point results as part of it.
Omniris hands over the underlying data in open formats: 3D models as GLB, point clouds as LAZ, Gaussian splats as PLY, maps and elevation models as GeoTIFF, terrain for CAD as DXF and LandXML, and findings and detections as GeoJSON, CSV or KML. A STAC bundle carries a checksum for each file where one is recorded, so an expert can show the files analysed are the ones delivered.
Claims and investigations turn on specifics: which shingle, which connection, which crack, seen by whom and when. Omniris gives three ways to record them, each tied to the evidence it is about.
On any photo or map, place a coloured pin or draw an outline and attach a note of up to 5,000 characters. The position is stored against the image's own pixels (and, on a map, its place on the ground), so the mark stays on the shingle it was put on. Anyone with access to the file can see and add annotations, and creating, editing or deleting one is written into the file's custody history.
A note can be attached to a file as a whole, typed or spoken. On the phone, a pilot or adjuster can record a voice memo standing in front of the damage; Omniris transcribes it automatically, and the audio stays with the note so the original words can be played back. If transcription fails, the memo is still kept. The file's owner and the pilot and client on a linked job can add notes; only the author can edit one, and a deleted note is hidden from view but kept, not erased.
A finding is a tracked issue at a place: a damaged slope, a failed flashing detail, a crack in a beam. It has a location, category, severity, status, assignee, comments and evidence images cut from the source imagery. Findings are made by hand on a photo, map, 3D model or panorama, from the phone in the field, or promoted from AI detections, thermal anomalies, LiDAR clearance points, annotations or change between flights. One made from a photo is placed on the ground from the camera's position, with a radius stating how uncertain that placement is.
Every change to a finding is sealed into its own chain of custody, and a finding certificate sets out the complete history: each step and who took it, the fingerprints of its evidence images, and each evidence source file exactly as it was when the evidence was attached. It is signed, timestamped and checked like every other Omniris certificate. See the Findings paper.
Many disputes are about change: what the roof looked like before the storm, what the slope looked like before the excavation next door, how much of the façade failed between two inspections. When the same site has been flown on two dates, the Site Navigator's Compare section puts them side by side:
Each comparison is graded for how well the two flights line up, because an offset between surveys looks like change. Results print as a PDF report or download as a spreadsheet, and the difference can be saved as a map layer with its own custody history. AI change lists objects that appeared or disappeared between two dates and records them as findings.
Omniris compares the imagery you have. A baseline is there if someone flew it, or if you bring one in: an older aerial survey, satellite scene or contractor's map already in your vault can be added to a site's timeline as dated imagery with its source named. The capture date is filled in from the file's acquisition details where it has them, never from the date the file was made, and otherwise entered by you. The file itself is not altered, so its fingerprint still stands as the evidence. A retention policy can keep a baseline for as long as it may be needed.
This is the core of what Omniris provides for claims and litigation work. The full design is in the Evidence integrity paper; this section explains it in terms of the rules above.
When a file arrives, Omniris calculates its SHA-256 fingerprint [28]; change one byte and the fingerprint changes completely. It keeps up to three: the content fingerprint as uploaded, the stored fingerprint of the bytes kept (different once a file is encrypted), and the delivered fingerprint of a download with location details written in. The Evidence copy download gives the file exactly as stored, with nothing added, so the copy you produce is the one the fingerprint describes. This is the "process of digital identification" Rule 902(14) has in mind [2].
Every file has its own chain of custody: uploads, views, downloads, shares and each use of one, moves, encryption and decryption, transfers, processing results, annotations, measurements, AI results and each person's correction of them, deletion and restoration, each with who acted, when, from where, and the file's fingerprints.
Events are sealed into a ledger that can only be added to. Every hour, or at once when a certificate is requested, new events are combined into one fingerprint (a Merkle tree) and sealed into an anchor that links to the one before it, is signed with the deployment's own key (ECDSA P-256) and is timestamped under RFC 3161 [26] by an independent authority (DigiCert, with Sectigo as the fallback). No event can then be altered, removed or reordered without breaking a fingerprint already signed and timestamped. Timestamps are renewed in batches before an authority's certificate expires, so the proof outlasts a long case.
Certificates are private by design: a certificate discloses who had custody, not the name of every viewer, and network details such as IP addresses go only into a separate appendix the file's owner can request.
A custody certificate is a signed PDF (PAdES [27]) that carries its own proof: the certificate data, a detached signature, the signing certificate, the timestamp authority's root certificate and a small checker program are all attached to it. It is saved automatically in a Certificates folder in your vault.
| Certificate | What it shows |
|---|---|
| File | The file's complete custody history, its content and stored fingerprints, and who had custody. Views appear as counts, under stand-in names |
| Folder manifest | Every file in a folder and its subfolders (up to 2,000), each with its latest proven event and fingerprints |
| Network appendix | For the file's owner: the network details recorded with each event |
| Disposition | Each file deleted at the end of a retention period, its last stored fingerprint and the sealed record of its deletion |
| Finding | A finding's complete history, who did each step, its evidence fingerprints and each evidence source file as it was when attached |
There are three ways to check a certificate, none of which needs an Omniris account:
For a production of many files, the custody events of your vault can also be exported as a spreadsheet (up to 10,000 rows per export), filtered by event type or by file.
Every deployment re-checks everything added to its custody record since the last check daily, and rebuilds the whole record from the beginning weekly, alerting the people who run it on any mismatch. Certificates never depend on this check, but it means a problem is found by the record's keepers first, not by an opposing expert.
Rule 37(e) asks for reasonable steps once litigation is reasonably foreseeable [8], and the Sedona Conference's commentary on legal holds sets out guidelines on when a hold is triggered, how it is carried out and documented, and when it is released [9]. In Omniris a hold is placed on what the dispute is about:
| Hold on | Covers |
|---|---|
| File | That file |
| Folder | Every file and subfolder in it, including anything added later or moved in |
| Job | The job's folders and delivered results, plus its operational record: mission plans, airspace authorizations, pre-flight checklists, flight logbook entries, tracked flights and flight paths, messages, attachments and invoices |
| Site | All imagery tagged with the site, including files tagged later |
| Account or organization | Everything it owns, including anything it gets later |
While something is held it cannot be edited, replaced by reprocessing, moved to another account, sent to the recycle bin or deleted, by anyone. Files already in the recycle bin when the hold is placed are restored. The hold is enforced in four independent places: the application, the storage layer, the database, and a locked copy of each held file in write-once storage (escrow), checked against its custody fingerprint as it is copied and refused if it does not match. The escrow lock can be extended but never shortened or removed, so held evidence survives even a compromised account. A red gavel marks each held file and folder in the vault.
A retention policy keeps every file in a folder or site for a set period, then either releases the files or deletes them permanently, with a signed disposition certificate. A legal hold always wins over a retention policy. The policy's creator and the data's owners are warned 90, 30 and 7 days before files reach the end of their retention, and a locked policy can never be removed or shortened. Retention keeps claim files for the period your record-keeping rules require; California, for example, expects claim data for the current year and the four preceding years to be retrievable [24].
Sensitive files, such as claimant interiors or privileged work product, can be protected with OpenTDF, an open standard for encrypted files that carry their own access rules (AES-256-GCM). Each opening is checked for that person and recorded before the contents leave, and the certificate gives the fingerprints of both the original and the encrypted copy. A held encrypted file's key is kept with the hold under a separate recovery key, so it can still be opened years later. See the Security paper.
Omniris provides verifiable records and the means to check them. Whether particular evidence is admissible, and what weight it carries, depends on the jurisdiction and the case. A Rule 902(14) certification is made by a qualified person; Omniris gives that person the fingerprints, the history and the tools to recompute them.
Carriers with drone teams fly their own claims; everyone else needs a pilot near the loss, quickly. On the Omniris cloud and the data center edition, the marketplace connects the two. (Marketplace, payments and payouts are not part of on-premises installations, where an organization's managers assign their own pilots.)
The location (find the site by address or pin, then draw the area to be flown, with no-fly zones marked), the deliverables (photos, video, orthomosaic, 3D model, thermal, multispectral, LiDAR, raw data), the dates and a point of contact. Numbered pins with notes ask for specific ground-level shots, such as "north elevation, gutter damage". Attachments can go with it.
Approved pilots whose travel radius covers the location are emailed, or you send the job straight to a pilot who has completed a job for you before. Until a pilot is assigned, browsing pilots see only the city and an approximate pin; the exact address and contacts stay private, which matters when the site is a claimant's home.
A pilot's Part 107 certificate is checked against the FAA's airmen registry when the pilot enters it. The registry does not show recurrent training, so a pilot whose certificate was issued more than 24 months ago is asked to upload the recurrent training certificate.
Pilots check the mission against FAA airspace, get the exact LAANC request to file (area, ceiling, time window) and record the authorization with the job, and keep a flight log. The Part 107 pre-flight checklist runs on the phone; once started it can be completed without signal, it files a snapshot of the airspace and the day's weather, and it cannot be deleted once complete. All of it belongs to the job, so a hold on the job keeps it with the imagery. See the Mission planning and Field app papers.
A frozen, read-only snapshot of the results lands in your vault, each file's custody history starting from where it came from. Accept it (which unlocks downloads) or send it back with a note.
Jobs can repeat on a schedule for sites that need documenting through a long dispute or repair. A pilot can also log a job for a client of their own, outside the marketplace, and send the results by a link from their own email.
Omniris checks the airspace and keeps the authorizations, checklists and logs, but the remote pilot remains responsible for the flight under Part 107 [25], for state rules on capturing images [21][22], and for requesting emergency waivers after a catastrophe [18].
Handing imagery to an insurer, opposing counsel, an expert or a mediator is itself an event worth recording. Each way of sharing leaves a record of what was shared, with whom and when; the Sharing and delivery paper covers them in full.
| Way to share | What the recipient gets | Fits |
|---|---|---|
| Snapshot (sharing code) | A frozen, read-only copy of a folder in their own vault, as it was when redeemed, with its own custody history. Nobody can change it | A production set or exchange of record |
| Read Only or Full Access (sharing code) | Live access to a folder from their own account | Co-counsel, the adjuster, a retained expert |
| Public link | A browser page, no account needed: view only or view and download; optional password; expiry never, 1, 7, 30 or 90 days. View-only photos are shown up to 4,096 pixels with embedded GPS removed | A mediator, a policyholder, a claims examiner |
| Public 3D link | The 3D model, straight away, in the viewer | A meeting or a deposition |
| Site Navigator link | A map of one site with chosen dates and layers, and the measuring and comparison tools | Before-and-after views for the other side |
Two cautions for a production. A public link, a public 3D link included, opens protected (encrypted) files for anyone holding it; for named recipients only, use a sharing code. And a still image saved from the 3D viewer is downloaded, not filed, so it has no custody history until you upload it.
The same evidence features run wherever Omniris runs. What changes is where the signing key and the locked escrow copies live, and whether the marketplace is available.
| Edition | Signing key | Locked escrow | Marketplace |
|---|---|---|---|
| Omniris cloud on Microsoft Azure | Azure Key Vault; the key never leaves it | A separate storage account with a locked retention period on every stored version | Yes |
| Omniris cloud on Google Cloud | Cloud KMS; the key never leaves it | A dedicated storage bucket with a locked retention period on every copy | Yes |
| Omniris data center edition | The server's TPM security chip, or a protected key file | Write-once storage in compliance mode, with its own login | Yes |
| On-premises | The server's TPM security chip, or a protected key file | Write-once storage in compliance mode, with its own login | No; your own pilots |
An on-premises installation suits a carrier, a law firm or an engineering firm that must keep claimant data and work product on its own servers. If such a site loses its connection, anchors are still made and wait in a queue; when it reconnects, the next timestamped anchor proves that everything sealed before it existed by then, and the certificate states that times during the offline period rest on the local clock.
Evidence is only as good as the claims made for it. Omniris states the boundaries of what it records and measures:
Hail bruising, damage beneath the shingles and anything else behind a finished surface still need hands-on inspection [16][17]. Omniris documents what the camera saw and keeps a defensible record of it, alongside the inspector's own findings.
Omniris vertical paper · October 2026.