Cognyra logo
Notes Knowledge · Rag Semantic Search

Cognyra

Cognyra is a Hungarian AI note-taking app and assistant that ingests YouTube videos, podcasts, PDFs, documents and web pages, transcribes and indexes them, then answers questions with cited sources while monitoring chosen channels for genuinely relevant new content.

Active GDPR compliant Free plan Freemium API available Verified by Guidaio
Overview

What is Cognyra?

Cognyra is an AI note-taking application and assistant published by the Hungarian software company SoundCam Productions Kft. It turns scattered media — YouTube videos, podcasts, PDFs, Office documents, web pages and audio recordings — into a single searchable knowledge base that answers questions in natural language and cites the passage each answer came from. The pipeline runs in three stages. You attach a source by pasting a link or uploading a file. Cognyra transcribes it using the ElevenLabs Scribe engine, applies speaker diarisation and timestamps, then builds embeddings and a vector index. You can then query the whole corpus conversationally, run semantic searches that match meaning rather than wording, or receive automatic summaries. Two capabilities set it apart. The first is retrieval-augmented generation with mandatory citation: every answer links back to the document and section it came from, using hybrid retrieval that combines OpenSearch lexical matching with dense vector search. The second is continuous monitoring — you nominate YouTube, Instagram, TikTok or Facebook channels and news sources, define a topic filter, and the system alerts you by email or Slack only when new material genuinely concerns that topic, typically five to fifteen minutes after publication. The vendor contrasts this with keyword tools, which cannot understand what a podcast actually discusses. The product is deliberately tuned for Hungarian: chunking accounts for Hungarian inflection, the embedding model is chosen to search Hungarian documents well, and transcription is claimed to be professional grade for Hungarian speech. The publisher presents this as the gap international tools leave. Alongside the individual use case, Cognyra is sold as a company platform with five named verticals: customer support, HR, legal, engineering, and marketing and PR. Business features include SAML 2.0 and OIDC single sign-on, role-based access control, per-document visibility, audit logging and just-in-time provisioning. It connects to SharePoint, Google Drive, Confluence, Notion, Slack and Microsoft Teams, and exposes a REST API with webhooks. Outputs go beyond text: the system generates five- to six-panel comic-strip summaries, diagrams and mind maps from long sources, and exports transcripts as SRT or DOCX. Data is held in AWS eu-central-1 in Frankfurt.

What it does

  • Answer questions across all your uploaded content, citing the exact source passage
  • Transcribe YouTube videos, podcasts, meetings and audio files with speaker recognition
  • Monitor chosen channels and news sources, alerting only on genuinely relevant topics
  • Search by meaning rather than keywords across thousands of documents at once
  • Deliver a daily morning digest summarising the channels you follow
  • Condense long sources into comic-strip panels, diagrams and mind maps
  • Export transcripts and data as SRT, DOCX, JSON or CSV
Audience

When to use Cognyra / When not to

A quick filter to help you decide if Cognyra is the right fit.

When to use Cognyra

  • Hungarian-speaking teams and companies that need an AI assistant which handles Hungarian inflection properly and keeps data inside the EU
  • Customer support organisations consolidating policies, product documentation and past tickets into one cited, searchable knowledge base
  • Researchers, doctoral candidates and university students who must extract quotable, timestamped material from lectures, podcasts and papers
  • Content creators, analysts and communications teams tracking many YouTube, podcast and social channels without watching all of them
  • Legal, HR and engineering departments searching contracts, internal policies and technical documentation with source-referenced answers

When not to use Cognyra

  • People who consume only one piece of new content a day — the vendor itself calls this overkill
  • Anyone wanting sub-second alerts on markets or sport: Cognyra notifies within five to fifteen minutes and says so plainly
  • Teams looking to republish monetised public content, since this is a personal knowledge base and not a content management system
  • Users who want video editing: the product transcribes and interprets footage but never cuts it
  • Non-Hungarian speakers who need a localised interface, as the website and product UI are Hungarian only
Get started

How to use Cognyra

A typical end-to-end flow, from setup to results.

  1. Register at the sign-up page with an email address or a Google account — no payment card is required
  2. Claim the 200 starter credits granted on registration and confirm your free monthly allowance
  3. Attach a first source by pasting a YouTube, Instagram or web link, or uploading a PDF or audio file
  4. Let Cognyra transcribe, apply speaker recognition, generate embeddings and index the material
  5. Ask questions in natural language and check the cited source passage behind each answer
  6. Run semantic searches across several documents at once when you cannot recall exact wording
  7. Set up a topic watch by nominating channels and defining the subject you care about
  8. Choose how alerts reach you — push, email or Slack — and enable the daily morning digest
  9. Connect business systems such as SharePoint, Drive, Confluence or Notion, which inherit existing permissions
  10. Export transcripts and data as SRT, DOCX, JSON or CSV whenever you need them elsewhere
Quick read

Pros & Cons

Pros

  • Every AI answer carries a verifiable citation, which materially lowers the risk of unnoticed hallucination
  • Genuine Hungarian optimisation — inflection-aware chunking and professional Hungarian transcription
  • Unambiguous EU hosting in AWS eu-central-1 with TLS 1.2+ in transit and AES-256 at rest
  • Contractual no-training commitment with the LLM suppliers, stated without conditions
  • Sub-processors named publicly — AWS, ElevenLabs, OpenAI and Anthropic — rather than hidden behind an NDA
  • A permanent free plan requiring no payment card, plus one-off credit packs for occasional spikes
  • Broad input coverage, from Office files and OCR images to social links and audio

Cons

  • The website and interface exist in Hungarian only — the English path returns a genuine 404, so there is no English mirror
  • No public API documentation despite a REST API and OpenAPI specification being advertised
  • No named security certification: neither ISO 27001 nor SOC 2 is claimed or evidenced
  • Retention periods are never quantified, being deferred to an unpublished internal policy
  • The terms of service publish neither a company registration number nor a VAT number, which is unusual for Hungarian terms
  • The Business plan's multi-user capability is still labelled as forthcoming while the plan is already sold
  • Credit top-ups require an active paid subscription, so free-plan users cannot simply buy more capacity
Pricing

Pricing & Plans

A permanent free plan is available and requires no payment card, providing 1,200 credits per month, roughly thirty minutes of video. The cheapest paid tier is Starter at 9.90 EUR per month. All published prices are net of tax: 27% Hungarian VAT is added at checkout for Hungarian buyers, while EU businesses holding a valid VAT number and non-EU buyers are charged 0% under reverse charge or the applicable rules. Billing runs through Stripe with NAV-compliant electronic invoicing, and a subscription can be cancelled at any time without charge, taking effect at the next billing date.

Plan 1
  • Ingyenes (Free) — 0 EUR per month — 1
  • 200 credits monthly
  • chat
  • YouTube and transcription
  • community support
Plan 3
  • Pro — 24.90 EUR per month net — 35
  • 000 credits monthly
  • professional transcription options
  • faster processing
Plan 4
  • Business — 79 EUR per month net — 120
  • 000 credits monthly
  • SLA-level support
  • multi-user announced as forthcoming
Custom enterprise quote — above 120,000 credits monthly or for regulated sectors
  • dedicated EU environment
  • own DPA
  • guaranteed SLA
  • SSO and RBAC
  • bespoke integrations
Plan 6
  • One-off credit packs — 5
  • 000 credits for 4.90 EUR
  • 20
  • 000 for 17.90 EUR
  • 60
  • 000 for 49 EUR
  • no expiry but an active subscription is required
Special offers — Permanent free plan with 1,200 credits per month and no payment card required · 200 credits granted on registration
Prices and plans listed above may evolve. Always check the official pricing page before subscribing.
Trust & Privacy

Data, GDPR & hosting

A consolidated view of how Cognyra handles your data.

GDPR overview

GDPR implementation is concrete and documented. The privacy notice is drafted as an Article 13 information notice and names its legal bases: contract performance for the service and account management, legitimate interest for security and abuse prevention, legal obligation for statutory duties, and consent for non-essential cookies and marketing. The competent supervisory authority is named as Hungary's NAIH. Data subject rights are listed in full — access, rectification, erasure, restriction, objection, portability and withdrawal of consent — and the vendor says they are exercisable from the account interface. A standard Data Processing Agreement is available to business customers with a sub-processor annex and thirty days' notice of changes, and a 72-hour breach notification process is described. Transfers outside the EEA, if any, rely on adequacy decisions or standard contractual clauses.

Who owns the data?

The data controller is SoundCam Productions Kft., registered at 2364 Ócsa, Üllői utca 48., Hungary, reachable at hello@cognyra.ai. The homepage states plainly that your data remains yours. Uploaded content is used exclusively to build your own knowledge base, and the terms confirm the service is a hosted SaaS for uploading, organising and AI-processing your material. The vendor commits contractually with its LLM suppliers that submitted content may not be used for model training. Users can export everything at any time and delete their account from the settings menu. Processors with access are named: AWS for infrastructure, plus ElevenLabs, OpenAI and Anthropic, and Google for optional sign-in.

Reuse rights

The terms describe the service as a cloud SaaS supporting upload, organisation and AI processing of the user's own data, and reserve trademarks, brand assets, software components and documentation to the provider. They set no restriction on the user reusing the material they uploaded or the outputs generated from it, and no licence is claimed over customer content beyond operating the service. Portability is explicit: data can be exported in JSON or CSV, transcripts as SRT or DOCX, and the account deleted on request. Acceptable-use limits do apply — unlawful, harmful, abusive or unauthorised data-harvesting use is forbidden, and the provider may restrict or suspend an account that breaches them. Note the vendor also states the product is not intended for republishing monetised public content.

Data retention & training

Retention summary
Cognyra publishes the principle but not the numbers. Personal data is kept for as long as the purpose requires and for any period imposed by law, with exact durations deferred to an internal retention policy that is not published. Google sign-in data is the one specific case: it is held for the lifetime of the account or until a deletion request is fulfilled. Cookies are described as either session-based, expiring when the browser closes, or persistent for a category-dependent period that is likewise not quantified. Users retain practical control regardless: the account and all its data can be exported and deleted at any time from the settings menu, and the vendor lists erasure among the data subject rights exercisable in one click.
Trains on customer data
No
Subprocessors disclosed
Yes
DPA available
Yes
GDPR contact

Hosting summary

All data is hosted in the AWS eu-central-1 region in Frankfurt, Germany, a location named consistently across four pages of the site. The vendor states repeatedly that data never leaves the European Union. Encryption is TLS 1.2 or above in transit and AES-256 at rest, with S3 buckets using SSE-KMS and both the relational database and the OpenSearch index encrypted. Operational resilience is described in concrete terms: daily automatic backups, point-in-time recovery over 35 days, cross-availability-zone redundancy, a recovery point objective under one hour and a recovery time objective under four hours. Enterprise customers may request a dedicated environment and, optionally, storage located in Hungary. Sub-processors are disclosed by name — AWS, ElevenLabs, OpenAI and Anthropic — with Google Ireland and Google LLC additionally involved where Google sign-in is used. Transfers outside the EEA, should they occur, are covered by adequacy decisions or standard contractual clauses.

Hosting countries
🇩🇩 Germany
Hosting regions
EU
Watch-outs

Things to keep in mind

Risks and trade-offs to weigh before adopting Cognyra.

  • Cited answers feel authoritative, which can quietly erode the habit of reading the source — the citation proves where text came from, never that the summary read it correctly
  • Delegating monitoring to a topic filter means you only ever see what the filter caught; relevant material phrased unexpectedly can disappear without you noticing the gap
  • Comic-strip and diagram digests are designed to let you skim rather than read, a convenience that becomes a comprehension loss when the underlying material is genuinely complex
  • Performance claims on the homepage — 99.4% transcription accuracy, sub-two-second answers, 32 speaker recognition — are published without method or source and should not be treated as measured guarantees
  • The customer support ROI figures are labelled by the vendor itself as an example calculation, not results achieved with real clients
  • Security rests on the vendor's own statements: no ISO 27001, SOC 2 or named auditor backs the encryption, retention and no-training claims
  • Uploading company documents, past tickets and recorded meetings concentrates sensitive material in one place, so account access and role permissions matter more than the tool's own safeguards
Setup

Setup & Integrations

Technical difficulty

Very low for an individual. Registration takes an email address or a Google account, needs no payment card and is claimed to complete in thirty seconds; the vendor explicitly targets non-technical users, describing a grandmother test on the homepage. A first integration is advertised as taking five minutes. Enterprise deployment is more involved, covering SAML or OIDC single sign-on, role-based access control and synchronisation with SharePoint, Drive, Confluence or Notion, with permissions inherited through the Microsoft Graph API. The vendor offers a 30-minute demo, a usage guide and a team onboarding workshop, and puts team rollout at one week.

Deployment

Web appAPISlack app

Integrations

Microsoft SharePoint Google Drive Atlassian Confluence Notion Slack Microsoft Teams YouTube Firecrawl Stripe Azure AD Okta Google Workspace Auth0 JumpCloud

Supported languages

HungarianEnglish
Company

Behind Cognyra

Company name
SoundCam Productions Kft.
Founded
28/09/2005
Country of origin
🇭🇺 Hungary
Headquarters
2364 Ócsa, Üllői utca 48., Hungary
UBO
INFORMATION_NOT_FOUND
UBO country
INFORMATION_NOT_FOUND
Domain registrar country
🇺🇸 United States
Legal contact
Support contact

Social

Official links

Resources

All the official URLs gathered for verification and reference.

Compare

Alternatives

Tools that compete with or complement Cognyra.

G Google AlertsC ChatGPT
FAQ

Frequently asked questions

Is Cognyra free to start?
Yes. The free plan is permanent, needs no payment card and includes 1,200 credits per month, about thirty minutes of video. A card is only required to move to a paid plan or to buy a one-off credit pack.
Where is my data stored?
In the AWS eu-central-1 region in Frankfurt, encrypted at rest with AES-256 and in transit with TLS 1.2 or above. The vendor states data never leaves the EU, and enterprise customers can request a dedicated environment or optional Hungarian storage.
Will my content be used to train AI models?
No. Cognyra says uploaded content is used solely to build your own knowledge base, and that its contracts with the LLM providers, named as OpenAI and Anthropic, operate in no-training mode and forbid using submitted text for model training.
What sources can I upload?
YouTube, Vimeo, Instagram, Facebook and TikTok links, any web page URL through Firecrawl, PDF, DOCX, XLSX, PPTX, HTML and Markdown documents, images via OCR, audio files in MP3, WAV, M4A and WEBA, and plain text.
Which languages does it handle?
The AI understands and answers in Hungarian, English and a claimed thirty-plus further languages, with transcription tuned for Hungarian speech. Note that the website and product interface themselves are Hungarian only.
How is this different from Google Alerts?
Google Alerts matches keywords in text and cannot interpret a podcast. Cognyra transcribes the full content and understands it semantically, so it notifies you only when material genuinely concerns your topic, and it supplies a timestamp.
How quickly are alerts delivered?
Typically five to fifteen minutes after new content appears. The vendor states explicitly that this is not a sub-second service and is unsuitable for stock market or sports betting alerts.
Is a Data Processing Agreement available?
Yes. A standard DPA can be signed with any business customer, with the sub-processor list supplied as an annex and thirty days' advance notice of updates. Bespoke amendments are reviewed by the vendor's legal team.
What happens when my credits run out?
New requests stop being processed, but your existing content and the search function keep working. You can add a one-off top-up pack, which requires an active subscription, or move to a higher plan.
Can I cancel and take my data with me?
Yes. Subscriptions can be cancelled at any time free of charge from the billing settings, effective at the next billing date. Data can be exported as JSON or CSV, transcripts as SRT or DOCX, and the account deleted on request.
Conclusion

Should you pick Cognyra?

Cognyra occupies a clear linguistic niche. Its central promise is not that it does something no other assistant does, but that it does it properly in Hungarian — inflection-aware retrieval and professional Hungarian transcription, in a market where international tools handle the language poorly. Around that core it layers two genuinely useful behaviours: answers that always cite their source, and semantic monitoring that understands what a podcast is about rather than matching keywords. The data posture is unusually explicit for a company of this size. EU-only hosting is named down to the AWS region, the no-training commitment is stated without conditions, sub-processors are listed publicly, and a DPA is offered to business customers. What is missing is independent verification: no ISO 27001, no SOC 2, no named auditor. The security claims are credible and specific, but they rest on the vendor's word alone. Two tensions deserve attention. The site addresses two different audiences with two different navigations — a personal knowledge base for individuals on one hand, an enterprise AI platform on the other — and a reader may reasonably wonder which product they are buying. And the publisher is a company registered in 2005 with a real trading history, while the product itself dates from 2024 onwards; the maturity of the company should not be read as maturity of the software. Practical reservations remain: the interface is Hungarian only, the advertised REST API has no public documentation, headline performance figures are published without method, and the Business plan is sold with a multi-user feature still described as forthcoming. For a Hungarian-speaking team that needs cited answers over its own material and EU data residency, the free plan makes the evaluation essentially costless. For everyone else, the language barrier is likely decisive.