BLUN Voice

Understand speech, transcribe it, and speak naturally

BLUN Voice connects AI speech recognition and natural speech output directly to the BLUN platform. The model can turn conversations and audio files into text, give structure to spoken content and read written answers back out as natural speech. People therefore no longer have to type everything or take notes by hand in order to work with an AI.

Voice is more than a dictation feature. It forms the speech layer for complete workflows: a user speaks, Voice recognises what was said, a text model understands the task, BLUN carries out the necessary steps and Voice reads the result back out. The result is assistants you operate the way you would hold a conversation.

The model supports both speech-to-text and text-to-speech, so speech can be handled in either direction. Voice can also be used in real time, which means answers can be processed and spoken while a conversation is still going on.

Usage by tier

Voice has no token context window comparable to King, Queen or Prince. On the free tier and on paid tiers, usage is therefore described in terms of speech capacity — for example available audio minutes, maximum file length, transcription volume, speech output or real-time use. The specific amounts have to be set out in the pricing overview.

Clean transcripts of conversations

Voice turns meetings, interviews, phone calls and voice notes into text you can work with. Instead of an unbroken run of individual words, the result is a structured transcript with sensible paragraphs and a readable arrangement.

Depending on the workflow used, speakers can also be separated, timings assigned and important passages marked. The transcript can then be passed automatically to Queen or King, so that decisions, tasks, summaries or complete documents come out of it.

Typical applications are:

• meetings and discussions
• interviews and customer conversations
• phone notes
• voice messages
• dictation
• training sessions and talks
• audio and video content
• spoken work instructions

Turning conversations into results

A transcript on its own saves only part of the work. Combined with the other BLUN models, Voice can turn spoken content straight into results you can use.

From a meeting, for instance, the following can be produced automatically:

• a short summary
• a list of every decision
• tasks with the people responsible
• the dates and deadlines mentioned
• open questions and risks
• a follow-up email
• an entry for project management

Voice supplies the spoken content for this. Queen works it up professionally, while King can analyse connections, dependencies and consequences in complex meetings.

Speak instead of typing

With Voice, users can dictate tasks, compose messages and operate BLUN by speaking. That is particularly useful on the move, in workshops, at appointments away from the office and anywhere a keyboard is impractical.

A spoken sentence such as “Turn this customer conversation into a quotation and remind me about the reply tomorrow” can be handed straight into an agent process. Voice recognises the speech, Queen or King understands the goal, and the BLUN platform carries out the steps involved.

Natural speech output

Voice can read text back out as clear, natural speech. That makes it possible to build applications that not only listen but also speak. Tone, speaking rate and the character of the voice can be matched to the case at hand.

Possible applications are:

• digital phone assistants
• voice-controlled product help
• reading out texts and documents
• audio versions of articles and guides
• low-barrier user interfaces
• spoken status messages
• interactive learning material
• voice agents for internal processes

Real-time voice agents

In a voice agent, speech recognition, model intelligence, tools and speech output come together into an ongoing conversation. The user does not have to wait for a complete audio file to be processed. The system can listen, respond, ask questions back and carry out actions.

An agent of this kind might check the status of an order, prepare an appointment, take down a support request or explain information from an internal system. Depending on the task, the actual thinking can be handled by Prince, Queen or King. Voice makes sure the whole interaction can take place by speaking.

Where Voice sits in the BLUN model family

Voice is the shared speech layer of the BLUN platform. The model brings together speech input, transcription, speech output and real-time interaction, and can work directly with King, Queen, Prince, agents and tools.

Typical tasks for Voice

• “Transcribe this customer conversation and separate the speakers.”
• “Turn the team meeting into a summary with decisions and tasks.”
• “Write this voice message up as a professional email.”
• “Read this guide out slowly and clearly.”
• “Build a voice assistant that can retrieve order information.”
• “Translate this voice message and give the answer back as speech.”

When Voice is the right choice

Use Voice when:

• audio or spoken language has to be processed,
• meetings and conversations should be documented automatically,
• users should be able to work with BLUN without a keyboard,
• an application has to be able to speak naturally,
• a real-time voice assistant is being built,
• spoken content should trigger actions directly,
• texts should be made available as audio.

For conversations that are critical in legal, medical or financial terms, transcripts and automatically derived decisions should be checked before they are used further. Background noise, indistinct speech and several people talking at once can all affect recognition.