PixLab ForensicInvestigate AI-generated and manipulated media
Media intelligence and forensic analysis,
in one workbench
PixLab Forensic helps you investigate AI-generated and manipulated media across images, video, audio and text. Inspect EXIF and GPS metadata, review waveforms and spectrograms, examine video playback, search source text and explicitly request connected still-image assessment from one browser-based workbench.
Local inspection does not upload the source. Connected image AI and online maps are blocked by default and run only when enabled and explicitly requested. See processing and storage details →
PixLab Forensic brings deepfake review, image inspection, video, audio, text, and project management into one browser-based workbench.
Inspect visual evidence
Zoom, metadata, ELA, hashes and recorded GPS.
Review recordings closely
Waveforms, spectrograms and dedicated video playback.
Trace text references
Search terms, URLs, emails, dates and source passages.
Request connected assessment
Still-image assessment and image questions, only when explicitly requested.
Dedicated media workspaces
Choose the source.
Keep the examination focused
The horizontal tab bar separates Images, Video, Audio and Text into focused workspaces, with How-to & FAQ and Manage alongside them. Move between open sources without rebuilding the investigation each time.
Move between sources without reopening them
Every file you open joins the session list. Each media workspace keeps one current source, and supported per-source editing, cursor and search state stays available while you switch.
The last primary tab is remembered when browser storage is available. Refreshing clears session sources unless you explicitly saved them into a local project.
A guided introduction and practical answers, shown first to new visitors.
Images
Inspect visible detail, EXIF/GPS metadata, local authenticity indicators and optional connected image assessment.
Video & Audio
Replay footage in a dedicated video player or listen to recordings alongside a waveform and local FFT spectrogram.
Text
Paste or import source text, search terms, inspect counts and surface candidate URLs, emails and date/time references.
Manage
Open Projects, Users & Profile, Settings, hash sets and App Data controls from one place.
Media intelligence in practice
Turn a claim into
specific checks
Start with a question the source can help you investigate. Separate what was recorded, what you can directly observe and what still needs independent verification.
QuestionExaminationUseful outcome
Submitted photograph
Does the recorded context agree with the account?
Inspect the image and available camera, date, software and GPS fields.
Identify details to corroborate and information that is missing.
Start with the source.
Build the examination from there
Open the media you already have, inspect what the file itself reveals, and use connected services only when they add useful context. Keep direct observations separate from interpretations, then preserve useful findings through notes, projects and reports.
PixLab Forensic workbench showing a source alongside inspection and analysis tools.
Open Images, Video, Audio or Text so the workbench presents the tools relevant to that material.
2
Open the source
Bring the material into the workbench
Choose a file, drop multiple supported files into the source panel or paste text directly into the Text workspace.
ImageVideoAudioText
3
Examine locally
Inspect what the source itself reveals
Review metadata, image details and ELA, replay recordings, inspect spectrograms or search and extract references from text.
4
Add connected context
Use connected analysis only when it helps
Explicitly request supported image assessment, image questions or location reasoning when additional context is useful.
Connected interpretations supplement the examination. They do not replace direct observations or independent verification.
5
Record the findings
Corroborate, add notes and export
Keep recorded facts, analyst observations and connected interpretations distinct, then preserve useful findings in notes or an exported report.
SourceObservationInterpretation
Source
Visual details
Inspect
Recorded context
Review
Indicators & gaps
Important: metadata can be altered or removed. ELA can also respond to ordinary editing and re-saving. Treat every value and indicator as information to corroborate.
Image diagnostics & EXIF
Read the capture details
behind an image
Review available EXIF metadata alongside a zoomable image to examine how recorded dates, device details and software tags relate to the claim you are checking. When coordinates are present, continue into location intelligence without treating metadata as unquestioned fact.
Image inspection · current and local
EXIF & capture details
Review camera, date, software and GPS fields when available.
Visual inspection
Pan and zoom into regions that require closer examination.
Error-Level Analysis
Use recompression differences as an investigative indicator, not proof.
Local fingerprints
Generate SHA-256, MD5 and perceptual dHash values locally.
Known-file lists
Import MD5 or SHA-256 lists and compare opened images locally.
Operator notes
Record observations, questions and missing context alongside the source.
Read recorded GPS coordinates locally, or explicitly ask connected image intelligence where a photograph may have been captured when no location is recorded. Keep recorded coordinates and inferred location suggestions clearly separated, and verify either against independent sources.
Recorded GPS stays recorded GPS
Read latitude and longitude from supported metadata without converting an inference into a source fact.
Open geographic context only when needed
With connected services enabled, choose Map, Split or Show on map to load the OpenFreeMap basemap through MapLibre, with 2D and globe presentation available. No map credential is required.
Treat AI location estimates as hypotheses
Ask about landmarks, signs or scene details through a configured image-analysis service, then verify the suggestion independently.
Recorded GPS coordinates can provide a research starting point when explicitly viewed on the map.
Recorded location
GPS metadata
Extracted from the image file when present
Inferred location
Connected AI estimate
Suggested from landmarks or visual details
Recording review
Locate the passage. Return to the moment
Audio and video use separate local workspaces for reviewing recordings. Navigate the source, return to the relevant passage, and keep the recording available while you compare it with other material. See processing and storage for the local-review boundary.
Current · Local
Audio analysis and waveform inspection
Open a browser-supported recording, listen while following its playback position and move through a waveform to revisit relevant passages.
Local playback and waveform navigation.
Local FFT spectrogram with linear and logarithmic frequency axes, click-to-seek and playback cursor.
Focused review of pauses, transitions and changes in level.
Playback pauses when you switch away from the workspace.
Planned: transcription, speaker diarization and synthetic-voice analysis. Waveforms and spectrograms do not establish speaker identity or prove an edit occurred.
Current · Local
Video examination and recording review
Use a dedicated player to pause, seek and revisit visible details while keeping the recording open during the session.
Native browser playback with supported audio track.
Pause and seek controls plus available duration and video dimensions.
Clear feedback when the browser cannot play an unsupported or damaged source.
Switching workspaces pauses playback while the source remains part of the current session.
Planned: frame extraction, separate soundtrack waveforms and automated video diagnostics.
Paste source text or import a TXT or Markdown file to search statements, inspect their size and surface candidate links, email addresses and date/time references. Current text inspection stays browser-local and does not send the source text to AI.
Text inspection · current and localTXT and Markdown · up to 5 MB
Search and jump to context
Move through term matches and click extracted candidates to select their exact passage in the source.
Surface references, do not silently trust them
URLs, emails and date/time patterns are candidates for independent review, not verified facts.
Keep the original while editing a working revision
View or restore the original at any time. Cursor, scroll and search state are preserved while you switch sources.
Extracted candidates are unverified. PDF parsing, OCR, entity recognition and connected text analysis are planned. Current text inspection does not determine whether writing is AI-generated.
AI-generated content & deepfake review
Assess AI-generated imagery.
Investigate every media type
Connected assessment currently applies to still images. For suspected synthetic material in other formats, use the audio and video review tools or text inspection while broader automated detection remains planned.
PixLab Forensic keeps local image evidence and explicitly requested connected AI assessment clearly separated during deepfake review.
Still-image assessment · available Audio detection · planned Video detection · planned AI-writing detection · planned
Explicit assessment
You decide when connected analysis runs.
Opening a source does not start AI analysis. The active image is submitted only when you explicitly request an assessment and a working analysis service is configured.
Corroborate locally
Compare the response with independent indicators.
Review the explanation and service attribution alongside ELA, SHA-256/MD5/dHash fingerprints, metadata and available C2PA Content Credentials claims.
Important boundary
Treat the result as evidence to check, not a verdict.
A displayed confidence value is reported by the model. It is not a validated forensic probability, authenticity score or certification.
AI-assisted image questions
Ask focused questions about the image in front of you.
Explore a scene, visible detail or possible location. When you submit a request, the configured service can use the active image preview and available metadata as context.
Scene summaries and specific visible-detail questions
Metadata-aware questions
Possible location reasoning
Responses remain associated with the active image
Example questions
“Which visible details should I examine more closely?”
“What do the recorded camera and software fields tell us?”
“Which location clues are visible, and which conclusions are estimates?”
“What additional information would help assess this claim?”
Opening an image does not start AI analysis automatically. Review each response against the source and independent information.
Local browser analysis & optional connected services
Know how each part of the examination
is processed
Core inspection happens in your browser. Connected AI and online maps remain separate and are used only when enabled and explicitly requested.
Browser-local
Inspect without source upload
Image inspection, EXIF and GPS extraction, ELA, hashing, supported audio and video review, and text examination run locally.
Explicit AI request
Connect only when requested
An image preview and request context are sent to the configured AI service only after you enable connected services and submit the request.
Explicit map opening
Online mapping stays separate
Recorded coordinates are extracted locally. Opening the map connects to the OpenFreeMap basemap through MapLibre when services are enabled.
Processing and storage at a glance
Workflow
What it provides
Processing & retention
Image & EXIF inspection
Image details and supported metadata fields
Local.
Source and results remain session-only unless explicitly saved.
Image authenticity indicators
ELA overlay plus SHA-256, MD5 and dHash fingerprints
Local.
ELA is an investigative indicator, not proof.
Known-file checks
Match or no-match against imported hash lists
Local.
Compared in the browser, not against an online database.
GPS extraction
Coordinates recorded in the source file
Local extraction.
Online mapping is a separate action.
Audio review
Playback, waveform navigation and FFT spectrogram
Local.
Browser processing for supported recordings.
Video review
Playback, seek controls, dimensions and duration
Local.
Browser playback depends on supported codecs.
Text examination
Counts, search, URLs, emails, dates and linked passages
Local.
Pasted or opened text clears on refresh unless retained in a saved project.
Session sources
Multiple sources with batch intake and per-source state
Session-only.
Switching sources preserves supported state during the session.
Local projects
Saved investigations with source files, hashes and notes
Explicitly retained.
Stored in this browser; project manifests do not contain media files.
Connected image AI
Answers, summaries and synthetic-media assessment
Explicit request.
Sends an image preview and request context to the configured service.
Interactive maps
Geographic context for recorded coordinates
Explicit opening.
Connects to OpenFreeMap through MapLibre when services are enabled.
Local preferences
Tab, theme, text size, layout, display name and service settings
Stored in the browser when available. Preference export excludes service configuration.
Session continuity
Open multiple sources, batch-import files, and move between them during the session. Refreshing clears unsaved session sources, pasted text, and findings, while projects you explicitly save remain available in this browser.
Manage brings local profile, project, hash-set, display and app-data controls together. The shared header keeps text size, theme, fullscreen and general settings within reach.
Local projects with explicit retention
Save source files, hashes, and notes locally when you save a project. Export/import manifests without media files or credentials.
Local profile & preferences
Save an optional display name, remember the last workspace, resize text from 100%–150%, switch themes, use fullscreen, and collapse panels.
Data & service controls
Import local MD5/SHA-256 hash sets, manage connected analysis and maps, export interface preferences, or reset them. Exports exclude source media, findings, and service configuration.
PixLab Forensic project and workspace management controls stored locally in the browser.
Local export
JSON evidence bundle
Hashes, recorded metadata, analysis parameters, operator notes, provenance notes and an explicit disclaimer.
Print output
Print-ready report
Produce a structured report and save it as PDF through the browser print dialog. Source files are not embedded.
Reports record the examination; they are not certificates of authenticity. Preference exports exclude source media, findings and service configuration.
Built around investigative questions
Different source types.
One careful examination process
Use the workspace that matches the evidence in front of you, develop observations and then compare those observations with other sources.
Journalists & fact-checkers
Examine submitted images, recordings and statements before drawing conclusions about a claim.
OSINT researchers
Use coordinates, visible clues and extracted text references as starting points for independent research.
Investigators
Review supported source material, note missing information and develop concrete follow-up questions.
Law-enforcement analysts
Use the workbench for initial examination within the organization's evidence-handling procedures.
Audio & video reviewers
Revisit important passages and compare what is visible or audible with the account of an event.
Synthetic-media investigators
Combine connected still-image assessment with metadata, ELA, hashes and Content Credentials while broader detection capabilities develop.
PixLab Forensic already supports direct source examination across several media types. The roadmap extends those capabilities with deeper authenticity checks, recording intelligence and cross-source analysis.
Roadmap previews are illustrative
Demonstration graphs do not represent relationships automatically derived from files you open.
Planned capabilities have no committed release dates.
Error-Level Analysis is available for re-save inspection. JPEG compression analysis and regional noise comparison remain planned.
Content Credentials
Partial
C2PA presence and supported claim extraction are available. Full manifest parsing and signature verification remain planned.
Synthetic-media detection
Partial
Connected still-image assessment is available. Broader validated detection across image, audio, video and writing remains planned.
Entity relationship analysis
Planned
Relationships between files through devices, dates, locations and other observations are planned.
Projects & cross-source analysis
Mixed
Local project retention, notes and manifest import/export are available. Wider search and richer cross-source analysis remain planned.
Recording intelligence
Expanding
Local spectrograms and structured reporting are available. Transcription, diarization and richer automated recording analysis remain planned.
FAQ
Practical answers for
media examination
Quick answers about image forensics, recordings, connected AI, local processing, saved projects and reports. Important boundaries are highlighted so you can scan them quickly.
PixLab Forensic is a browser-based media intelligence and forensic-analysis workbench for images, video, audio and text. Use local inspection tools first, then explicitly request connected image AI or maps when needed.
What can media intelligence and diagnostics tell me?
They help you develop observations to verify: recorded metadata, visible image details, audio patterns, video moments and text references. They can also reveal missing context, but they do not prove a claim on their own. See analysis in practice.
How do I get started, and why does How-to & FAQ open first?
The guide opens first for new visitors. Choose Images, Video, Audio or Text, then open a supported file or paste text. Files from the shared source panel are routed to the appropriate workspace. See workspaces.
Does PixLab Forensic remember my last workspace?
Yes, when browser storage is available. The last primary tab can be restored, but unsaved session sources and findings clear on refresh. Explicitly saved local projects are separate. See processing & storage.
How can I make the text larger?
Use Aa (Text size) in the header or Manage > Settings. Choose 100%, 115%, 130% or 150%; the preference is remembered locally when storage is available.
02
Images, metadata & location
How do I read EXIF metadata from an image online?
Open the image in Images and review the Metadata panel. Supported fields can include camera make/model, capture date, software tags and GPS. Extraction runs locally and does not trigger AI. See Image diagnostics & EXIF.
Can I examine a photo without EXIF metadata?
Yes. You can still pan, zoom, use local image indicators and optionally request connected assessment. Missing EXIF does not prove authenticity, and metadata tools cannot recover fields that are absent.
Can I find where and when a photograph was taken?
PixLab Forensic can read recorded dates and GPS coordinates when they exist. Recorded values can be edited, so corroborate them. The optional map and AI location estimate are separate actions. See location intelligence.
Can PixLab Forensic estimate an image's location without GPS?
Yes, through an explicit connected image question. The service can reason from landmarks, signs and visible details, but the result is only a hypothesis and does not recover missing GPS.
Are image tampering analysis and Content Credentials available?
Current tools include visual review, metadata, ELA and C2PA presence/claim extraction. Deeper JPEG/noise diagnostics and full C2PA manifest/signature verification are planned. See the roadmap.
Can I check whether an image has been edited?
You can review ELA, metadata and hashes. ELA is an indicator, not proof of editing; ordinary re-saving can create similar differences. Exact hashes identify a file, while dHash helps compare perceptual similarity. See image diagnostics.
03
AI, deepfakes & media analysis
How does PixLab Forensic support deepfake detection?
For still images, an explicit connected assessment can return an explanation and reported signals. Compare that with ELA, hashes, metadata and available C2PA claims. Audio, video and text currently rely on local review tools; broader automated detection is planned. See AI / deepfake analysis.
Can it tell whether text, audio or video is AI-generated?
Not currently. Connected AI assessment applies to still images. Audio and Video provide review tools, and Text provides local search/reference extraction. Automated audio, video and AI-writing detection are planned.
Can I ask AI questions about an image?
Yes, when connected AI is enabled and configured. Ask about visible content, scene context, metadata or possible location. Each request is submitted explicitly and should be checked against the source.
What can I learn from audio waveform analysis?
Use the waveform and local FFT spectrogram to locate passages, pauses, transitions and level changes. Click the spectrogram to seek and follow the playback cursor. They do not identify speakers or prove an edit. See audio analysis.
What video analysis tools are available?
Current video review provides playback, pause/seek, duration and available dimensions. Frame extraction, separate soundtrack waveforms and automated video diagnostics are planned. See video review.
What can I analyze in the Text tab?
Paste text or import TXT/Markdown up to 5 MB. Current tools include:
Word, character and line counts.
Case-insensitive search and click-to-passage navigation.
URL, email and date/time candidates plus a restorable original.
Text stays local. PDF parsing, OCR, entity recognition and connected text analysis are planned. See text analysis.
04
Files, privacy & session behavior
Which media and text formats can I open?
Image display, metadata extraction, audio decoding and video playback depend on browser/library support. Text accepts TXT, Markdown and pasted text up to 5 MB. PDF and Word are not supported by the current Text workspace.
Does examining a file upload it to a server?
Core inspection does not upload the source. Connected image AI can send a preview and request context only when explicitly submitted; opening the online map creates a separate map connection. See processing & storage.
When does the workbench connect to the map service?
Only after connected services are enabled and you choose Map, Split or Show on map. Opening a GPS-tagged image alone does not connect to OpenFreeMap; local GPS extraction still works while connected services are blocked.
Are my files, text and findings saved after a refresh?
Unsaved session material clears on refresh. Interface preferences can persist locally, and projects you explicitly save retain their source files, hashes and notes in this browser. Review local projects and reporting → Reopening a saved project into the workbench is planned.
Can I keep several sources open at once?
Yes. Open or batch-intake several sources, then switch among them without reopening. Each media workspace keeps one current source and preserves supported per-source state during the session. Cross-file comparison and similarity search are planned.
05
Projects, reports & trust boundaries
What can I manage in Users & Profile and App Data?
Optional local display name and interface preferences.
Local projects and manifest import/export.
MD5/SHA-256 hash sets and connected-service controls.
The local profile is not an account. Preference export excludes media, findings and service configuration. See projects & app data.
Can I export a forensic report?
Yes. Report can produce:
JSON evidence bundle with hashes, metadata, parameters and notes.
Print-ready report that can be saved as PDF through the browser.
Source files are not embedded, and the report is not an authenticity certificate.
Does the entity graph show relationships from my files?
Not yet. The graph preview currently uses demonstration data. Automatic relationships derived from your opened sources are planned. See the roadmap.
Can the workbench prove that media is authentic?
No single metadata field, visual signal or AI response proves authenticity. Treat the workbench as a way to develop observations and questions, then corroborate them with source context and independent evidence.
Is PixLab Forensic free, and do I need an account?
The browser workbench is free and requires no account or desktop installation. Optional connected AI needs a configured analysis service; online maps need no credential. Connected services may have separate costs.
Is PixLab Forensic useful for OSINT, journalism and investigations?
Yes. It supports initial examination for journalists, fact-checkers, OSINT researchers, investigators and recording reviewers. Use it to inspect recorded context and develop follow-up checks; deeper case management remains planned.
PixLab Forensic · Free browser workbench
Give the source
a closer examination
Read the context recorded in an image. Revisit a moment in a recording. Find references in source text. PixLab Forensic brings those tasks into one focused workbench, with optional connected image intelligence when you need it.
