YOUR PRIVATE WORKSPACE
Bring the material.
Keep links, documents, photos, and videos together for an investigation.
Return to RealityEngine · App home
Choose a supported file or a public link. Saving and local review do not start research.
Add material
Checking private storage…
Sign in to Dark NovaReview a link before investigating
Write your own question, or read a limited page-text excerpt and review it first. Continue carries your reviewed words and reference into RealityEngine; explicit Load and Run are still required.
Read page text
This sends the public HTTPS URL to Tavily to read a limited text excerpt. It can use provider usage and a daily reading allowance, but does not use a Deep Scan credit or start an investigation. Page reading can fail or omit content, including pages that require sign-in, block extraction or depend on scripts. Returned text is unverified; compare it with the original page.
Checking page reading availability…
This is a limited text preview, not full-page analysis or verified evidence. Keep only relevant text and remove anything you do not want sent in a later investigation.
0 / 6,000 draft characters.
Review material text
Text is read on this device. This review does not upload the document, verify its claims, or start an investigation. TXT and selectable PDF text load locally. For scanned PDF pages or a photo, select Read English text. Recognition reads printed English and may misread words, numbers and reading order; compare the result against the original. This does not interpret pictures or videos.
Continue carries an editable draft into the investigation form in this browser tab. It expires after 10 minutes. Load the draft there, review your scan choice, then select Run only when you are ready. Research may use credits and sends the submitted claim to research providers.
The text reader runs on your device. First use downloads its files from Dark Nova and may use mobile data. It previews one photo or only the first three PDF pages, with up to 6,000 characters, and stops after 90 seconds. You can cancel. Your selected file stays on this device unless you select Save privately.
Review three video frames
Select Preview three frames to sample approximately 10%, 50% and 90% of the selected clip on this device. Samples can miss events between frames. This preview does not listen to audio, interpret scenes or verify authenticity. The video and frames stay on this device unless you separately select Save privately for the video.
MP4 or WebM, up to 16 MiB and 5 minutes. Some codecs cannot be read by your browser. Preview stops after 30 seconds and can be canceled. Large frames are reduced for display.
Write only what you can observe in each sample. These notes remain unverified input for a later claim investigation. Each field allows up to 1,200 characters.
Continue carries your notes and approximate timestamps as a ten-minute draft in this browser tab. It sends no frames and starts no research. The investigation form requires explicit Load and Run; running the submitted claim may use credits and send your notes to research providers.
Review a short clip with sound
Prepare an MP4 or WebM up to 60 seconds and 8 MiB locally. Some containers and codecs cannot be prepared. Saving privately remains a separate action.
The AI review sends the whole clip, including visible faces, voices and metadata inside the file, plus an extracted soundtrack where present, to Cloudflare Workers AI. Video is sampled at 2 frames per second and can miss events. Speech recognition can mishear or invent words. The assessment and speech text are unverified; this does not establish identity or authenticity. This can incur provider usage and use one review allowance per 24-hour window, without debiting a Deep Scan credit.
Compare the words and assessment against the clip. Edit the speech text and question, and remove unsupported or private details. These probabilities describe model choices, not the truth of the claim. Dark Nova does not retain the clip or review on its server through this action; request fences prevent repeated submissions. Saving the original privately is separate.
0 / 6,000 draft characters.
Continue makes an editable ten-minute draft in this browser tab. Explicit Load and Run are still required in RealityEngine; research may use credits and send the submitted words to research providers. No clip or soundtrack is carried into that investigation.
Describe a photo
First prepare a reduced JPEG on this device. JPG, PNG or still WebP up to 16 MiB and 16 million pixels. Preparation stops after 20 seconds and can be canceled. It removes the original filename, EXIF location metadata and comments from the copy. Visible faces, addresses and other details in the picture remain visible. Review what you share.
Checking AI description availability…
The owner pilot needs one activation before a photo can be sent. Review the Llama 3.2 Community License and Acceptable Use Policy.
Your original approval and failed attempt remain recorded. A recovery makes one additional activation request to Cloudflare Workers AI with at most one output token. It may incur AI usage charges even if the original request already registered the terms. It does not send a photo, use research credits or start research. It cannot run automatically or repeat after a lost response.
Sending this prepared copy gives Cloudflare Workers AI access to its visible content for a short AI description using Llama 3.2 Vision. Built with Llama. Dark Nova keeps the description privately for up to 24 hours so you can recover a lost response; scheduled cleanup can take longer during an interruption. This description step does not save the photo in your materials, use research credits or start an investigation. The AI provider may incur usage charges.
An AI description can invent or miss details. Compare it against the photo, correct it and remove unsupported claims. Continue creates a ten-minute editable research draft in this tab. Explicit Load and Run are still required; research may use credits and send the claim to research providers.
Your materials
Received material is not verified evidence and is not automatically analyzed. TXT and selectable PDF text can be reviewed locally before starting a separate claim investigation. English text in scanned PDFs and photos can be recognized locally after selecting Read English text. Selected videos can be reviewed as three local frame samples with your own notes. AI photo description is available only where an owner pilot is explicitly enabled. Links support an explicit, consented page-text reading where available; the excerpt remains unverified and research requires a separate Load and Run. Selected short clips also support an explicitly consented AI review with speech text and a question about sampled footage, where available. Review outputs remain unverified.
Files are stored as supplied and are not malware-scanned. Text review reads the file you select on this device; it does not open previously saved materials. Deletion removes the stored object; a limited deletion record is retained for up to 30 days to prevent an interrupted transfer from restoring it.