kayetana
Solutions for media & publishing

Research with AI. Sources stay in-house.

Interviews, background conversations, archives and unpublished manuscripts: in newsrooms and publishing houses, much of it is confidential, and some of it protects sources. kayetana helps you transcribe, search and organize, on your own server.

kayetana.local
Audio transcription with timestamps
Audio

Transcribe interviews without uploading them.

Upload recordings as MP3, M4A, WAV, AIFF or FLAC up to 100 MB. kayetana transcribes them with Whisper large-v3-turbo, complete with timestamps. You can find a quote in the recording quickly, and the conversation stays on your server.

  • Timestamps to jump back in
  • Selectable language
  • Also via API with transcribe
Use cases

Where kayetana helps in newsrooms and publishing.

Query your archive

Load past articles and get answers with a source for every sentence.

Tag articles

Classify articles by desk, topic or region, with a confidence score per label.

Summarize manuscripts

Get summaries of proposals and manuscripts before editors read them.

Find related stories

Rank articles by how close they are in content with embed and rerank.

Draft variations

Get suggestions for teasers and headlines in your house style, with team templates.

Sort reader mail

Route letters by topic so they reach the right desk faster.

Answers from documents

Your archive answers with sources.

Load articles, dossiers and research files as PDF, Word or HTML. kayetana answers only from these texts and adds a source number after every sentence. That way, you check every statement against the original before you use it.

  • A source for every sentence
  • Several documents in one request
  • Also via API with citations
Answer with a source for every sentence
Getting started

How your newsroom gets going.

Pick a pilot

For example, a desk that conducts a lot of interviews.

Set up the server

Install kayetana on a Mac Studio on your network.

Invite your team

Invite editors by link and assign roles.

Set retention

Define how long the usage log and datasets are kept.

See kayetana with a sample recording.

In the demo, we transcribe a sample interview and query a sample archive with citations.

FAQ

Frequently asked questions

Do interviews and sources stay confidential?

Recordings are processed on your server. Prompts and responses are never stored, and once the models are loaded, no internet is needed.

How long can recordings be?

Each file can be up to 100 MB. Supported formats include MP3, M4A, WAV, AIFF and FLAC.

Does kayetana write articles?

kayetana drafts text like teasers or summaries. You review and approve it yourself.

Can I connect kayetana to our content management system?

Yes, through the REST API under /v2 with API keys. The playground shows sample code for Python, TypeScript and curl.