telkit

Use case

Call recording with a transcript, ready for your notes

A salesperson taking notes during a call listens less well. A recording nobody plays back is just a liability. With the call as text, the notes are quick to write.

Last reviewed 14 October 2026

The problem

After a customer call someone has to write down what was agreed. Written from memory, the details go missing. Scrolling through a recording takes as long as the call itself.

telkit records the call and writes it out when it ends, with who said what and when. You write the notes yourself, or let your language model do it, from text you can search.

The services that do the job

Voice for the calls, transcription for recordings you already have.
  • A phone in the browserCall the customer from the dashboard with recording and text switched on. The call goes out from a Norwegian number the company rents through telkit.
  • RecordingSwitch on recording for calls you place, and for calls coming in to a number.
  • Text after the callWhen a call you placed ends, the recording is written out with who spoke. The text sits with the call in the dashboard.
  • Transcribing your own recordingsIf you have a meeting recording, send the file or a link to the transcription API. The answer comes as JSON with timestamps, SRT, VTT or plain text.

Example flows

From a call or a meeting to notes somebody actually reads.

The sales call

  1. The salesperson calls the customer from the dashboard with recording and text on.
  2. She says at the start that the call is recorded, and why.
  3. After the call she reads the text with the call and writes the notes into the CRM.

A video meeting recording

  1. The meeting was recorded as WAV. Your system sends a link to the file to the transcription API.
  2. transcription.job.completed says the text is ready, and how long the audio was.
  3. Your system fetches the text, gives it to the language model you use, and stores the notes with the customer.

Customer service recording incoming calls

  1. The number is set to «Hilsen, så viderekobling» (a greeting, then forward), with recording on.
  2. The greeting says the call is recorded before the caller is put through to an agent.
  3. The recording sits with the call in the dashboard for the people with access.

What the request looks like

A transcription job from a link, the event when it is done, and the text as subtitles.
POST /v1/transcription/jobsbash
curl https://api.telkit.no/v1/transcription/jobs \  -H "Authorization: Bearer $TELKIT_API_KEY" \  -H "Idempotency-Key: kundemote-2026-10-14" \  -H "Content-Type: application/json" \  -d '{    "profile": "salg",    "source": {      "kind": "url",      "url": "https://eksempel.no/opptak/kundemote-2026-10-14.wav"    },    "language": "no",    "diarize": true,    "name": "Kundemøte 14. oktober"  }'
Without language, telkit detects it. diarize tells the speakers apart and is on unless you switch it off.
Webhook: transcription.job.completedjson
{  "object": "event",  "id": "evt_…",  "type": "transcription.job.completed",  "api_version": "1",  "occurred_at": "2026-10-14T08:00:03Z",  "data": {    "eventId": "evt_…",    "mode": "live",    "profileId": "…",    "jobId": "…",    "seconds": 1260,    "language": "no",    "priceOre": 126  }}
seconds is the length telkit measured on its own copy of the file. The price is per started audio minute.
GET /v1/transcription/jobs/{id}/resultbash
curl "https://api.telkit.no/v1/transcription/jobs/{id}/result?format=srt" \  -H "Authorization: Bearer $TELKIT_API_KEY"
format can be json (text and timestamps per speaker), srt, vtt or txt.

What to watch out for

An audio recording of a call is personal data about everyone who speaks.
  • The Norwegian Data Protection Authority (Datatilsynet) writes that businesses as a rule must ask for consent before they record a call with a customer. Inform people before the recording starts.
  • Section 205 of the Norwegian Penal Code covers secret recordings of conversations between others that you do not take part in yourself. The recording must belong to the call you are in.
  • Decide how long recordings and text are kept before the first recording. You delete a transcription job with DELETE, which removes both the audio and the text.
  • Separate who may hear the recording from who only needs the notes. Most people only need the notes.
  • Speech to text runs through a subprocessor. See who, and where the data is processed.Subprocessors

What it costs

Recording and text of a call are paid per started minute of the call. Transcribing a file is paid per started audio minute.
Outbound call, per minute
NOK 0.60
Recording, per minute of the call
NOK 0.20
Text of the call, per minute
NOK 0.12
File transcription, per audio minute
NOK 0.06

A 20-minute sales call with recording and text costs NOK 18.40. A 60-minute meeting recording costs NOK 3.60 to transcribe. Both excl. VAT.

Prices exclude VAT.

Questions about call recording and transcripts

Do I need consent before I record a customer call?

Datatilsynet writes that businesses as a rule must ask the customer for consent before recording the call, and inform them about the recording.

Does telkit write the notes?

No. telkit gives you the text with who said what. You, or your language model, write the notes from the text.

Can incoming calls be recorded?

Yes, recording is switched on per number. Text after the call is made for calls you place yourself, from the browser or the API.

Which languages does transcription understand?

You can ask for Norwegian or English, or let telkit detect the language. The answer says which language it heard.

What is diarisation?

Splitting the text by who is speaking. Each piece gets a speaker and a start and end time in seconds.

Read more about the services

Record your first call

Register your company, call yourself from the dashboard with recording and text on, and read the text afterwards.