AgentykLM: ten tiers

Reads a message and answers your questions about it.

Emails, chat messages, order notes, documents and data records arrive all day, and someone has to decide what each one is, how urgent it is and where it goes. Agentyk Decision does that reading for you.

Send the text with up to 16 questions in one request. A question can be a yes or no, a choice from your own options, or a score on your own scale. Every answer comes back clear, with probabilities, so you set your own thresholds and send each message where it belongs.

• Up to 16 questions in one request
• Yes or no, a choice from your own options, or a score on your own scale
• Probabilities with every answer, so you choose the thresholds
• Available as a closed beta: ask us for access

Turns a recording into clean text.

Meetings, interviews, calls and voice memos all hold something you need, but getting it out means listening back, pausing, rewinding and typing it all by hand.

Send an audio or video file to Agentyk Scribe and get the transcript back. Every passage comes with its timestamps, so you can jump straight to the moment it was said. It works in many languages and writes each recording down in the language it was spoken. Expecting only certain languages? Tell it which ones to allow, and a recording that does not match can be flagged instead of transcribed.

• Audio or video in, a transcript out
• Timestamps on every passage
• Many languages, each written down in the language it was spoken
• Allowed languages, with a mismatch flagged on request
• The audio is not kept after transcription

Writes it down while people talk.

In a call or a meeting everyone talks at once, and the notes fall behind. By the time one point is written down, three more have gone by, and afterwards nobody is sure who said what.

Agentyk Scribe Live streams the text back while the audio is still being spoken. Words appear as they are said and settle into clean text a moment later. Each line can show who is speaking, for up to eight speakers, and when the session stops it can check the speaker labels once more across the whole session. In AgentykCloud you can use the microphone and also capture a tab or a call.

• Text while people talk, settling into clean text
• Speaker labels for up to eight speakers
• An optional re-check of every speaker turn when you stop
• Microphone, a browser tab or a call in AgentykCloud

Live captions in up to five languages while people talk.

In a meeting with colleagues from across Europe, not everyone follows every language. Some catch a word here and there, others wait for someone to sum it up later, and by then the moment has passed.

Live translation is an option of Agentyk Scribe Live. While someone speaks, each finished sentence is transcribed and then translated into up to five languages at once. The captions follow about two seconds behind the speaker, each translation arrives about 0.4 seconds after its sentence is written down, and the speaker labels are carried along. Use it in AgentykCloud, or add the translation option with your target languages to the start message of the live transcription connection.

• Up to five target languages at once
• 24 languages, among them English, Dutch, German, French, Spanish and Italian
• Captions about two seconds behind the speaker
• The speaker label on every translation
• Hosted in the EU, the AI runs on our own servers in the Netherlands, and translations are not stored
• Available on every Scribe Live account: €0.002 per audio minute per language

Turns text into natural Dutch speech.

Order updates, directions and answers to questions all need a voice, and a stiff, robotic one puts people off. Many voices also stumble over Dutch.

Agentyk Voice turns text into natural Dutch speech, and speaks English too. Send the text and get the spoken audio back. In a live conversation it streams the speech, so it starts talking while the rest of the text is still on its way, and when someone interrupts, your app can stop it at once. The express model is lighter and lets you set the speed.

• Dutch and English
• Spoken audio back as a WAV file
• Streaming speech for live conversations
• Stops at once when someone interrupts
• An express model with adjustable speed

Transcription tuned for Urdu.

Calls, interviews and voice notes in Urdu are hard to transcribe well, so the text often comes back garbled and someone ends up typing it out by hand.

Agentyk Scribe Urdu is tuned for Urdu. Send an audio or video file and get the transcript back in Urdu script, written from right to left as it is read, with timestamps on every passage. It is used just like Agentyk Scribe: the same request and the same options, with the Urdu model chosen. Its language detection favours Urdu, so use Agentyk Scribe when you need to find out which language a recording is in.

• Tuned for Urdu
• Transcripts in Urdu script, right to left
• Timestamps on every passage
• The same request and options as Agentyk Scribe