Skip to Test Lab
Demo boundary · fictional data only · no government connection
ChallanSakshiचालान साक्षीby sh1rsDemo

TEN CASES. ONE EVIDENCE ENGINE.

Synthetic Evidence Test Lab

Run the cases. Inspect the reasoning. Change a fact and see what changes.

Compare source-linked observations, test uncertainty, and follow each result to a prepared next step. Every case stays open for inspection.

Full challan walkthrough
Or follow the 90-second guided proof

SYNTHETIC 90-SECOND PROOF

Looking for the 10-case Test Lab? Open all ten cases ↓

Does this fictional image show the same vehicle as the record?

See one complete evidence-to-handoff loop, including a correction that clears the earlier result.

FICTIONAL VEHICLE RECORDBlue Honda Activa 6G · Two-wheeler
Synthetic photo of a blue scooter; plate KA01AB3317SYNTHETIC TEST PHOTO
Plate detail · same photo
Enlarged original pixels: Synthetic photo of a blue scooter; plate KA01AB3317
FICTIONAL ENFORCEMENT IMAGEWhite Maruti Swift · Four-wheeler
Synthetic photo of a white hatchback; plate KA01AB3317SYNTHETIC TEST PHOTO
Plate detail · same photo
Enlarged original pixels: Synthetic photo of a white hatchback; plate KA01AB3317

Synthetic photos with visible plates · pre-authored observations · no live model reading

AI extractsSource-linked observations

Rules compareBounded deterministic outcomes

Citizen controlsConfirmation and every handoff

Synthetic demonstration data. No real record, upload, government connection, filing, payment, or decision.

RUNTIME EVALUATION

Synthetic Evidence Test Lab

One engine, not ten canned conclusions.

Press Run all to calculate every actual outcome.

SELECTED TEST VECTOR · CASE-02-REGISTRATION-CONFLICT

One plate character differs

The record ends in 3317 while the clear enforcement plate ends in 3817.

Human confirmation requiredEngine Locked
  1. 1Evidence
  2. 2Explain
  3. 3Verify
  4. 4Act

01 · EVIDENCE

Review every source value before comparison

Bundled fictional test vector
See the original synthetic caseSource cards stay unchanged when you edit the observations below.
01 · FICTIONAL RECORD

Fictional challan

Reference
CS-LAB-2026-001
Registration
KA 01 AB 3317
Allegation
Riding without a protective helmet
Amount
INR 1000
Recorded time
2026-08-20 11:08 IST
Recorded place
Model Avenue, Pilot City
Authored test record · not government-issued
02 · FICTIONAL RECORD

Fictional vehicle record

Registration
KA 01 AB 3317
Vehicle
Scooter
Make / model
Honda Activa 6G
Colour
Blue
Authored reference · not proof of ownership
03 · SYNTHETIC PHOTO

Fictional image observations

Synthetic photo of a blue scooter; plate KA01AB3817SYNTHETIC TEST PHOTO
Plate detail · same photo
Enlarged original pixels: Synthetic photo of a blue scooter; plate KA01AB3817
Synthetic test photo · pre-authored observations, not a live AI reading.
Authored plate observation
KA01AB3817
Test time / overlay
20 Aug 2026 11:08 IST
Test location / overlay
Model Avenue Pilot City

RECORD FACTS

Record facts used in the action pack

These facts are not compared with the image, but they are included in the downloadable summary. Review them before confirming.

Challan reference
Issue date
Alleged offence
Amount
Registration
Vehicle type
Colour
Make / model
Timestamp
Location
Can the alleged offence be assessed from this image?

This asks only whether the relevant visual area exists. It does not ask AI whether an offence occurred.

Analysis provenance: Pre-authored bundled synthetic observations · no model call in this proof.

02 · EXPLAIN

The engine shows its rule trace

Primary Clear registration or recognized vehicle-type differences can support a potential evidence discrepancy.

Context One colour, model, time, or location difference alone asks for corroboration.

Abstain Partial, unclear, not-visible, or low-confidence values remain inconclusive—even if a guessed value exists.

Control AI observations never choose the official pathway. Deterministic rules do that only after your review.

03 · VERIFY

Confirm the current observations

Check all four record facts, six comparison groups, every source/confidence label, and offence visibility. Editing anything after comparison clears the previous confirmation and action.

Review the source observations before comparing them.

04 · ACT

One safe next step, after confirmation

Action is locked

Review and confirm the current source observations first. No stale result is reused.

CUSTOM TEST INPUT

Prove the rules are not tied to a fixture ID

Browser-local manual mode
The selected image is not transmitted by this lab.

Use a synthetic test image only; ChallanSakshi cannot verify its provenance. Its bytes stay in this browser tab and are not uploaded, stored, or sent to AI, OpenAI, ChallanSakshi’s server, or an authority. Copying or downloading a later text pack is a separate action you control. Do not choose a real challan, RC, face, plate, address, QR code, or location-bearing file.

Why does the public lab use manual observations?

Public live uploads remain off. The codebase includes a policy-limited, feature-flagged OpenAI extraction route for controlled testing, but arbitrary public uploads require verified access and abuse controls, spend limits, and metadata-stripping first. A person’s assertion that a file is synthetic cannot prove its provenance.

Evidence assistant, never judge.

Nothing is filed, paid, authenticated, or sent to a government system. The designated authority—not ChallanSakshi or an AI model—decides any official outcome.