Audio Models — ElevenLabs, Sonauto, OpenAI
Audio models in Krater Krater covers the full audio stack: text to speech, live voice conversations, music, sound effects, voice changing and transcription. Audio runs on ElevenLabs, Sonauto and Op...
Audio models in Krater
Krater covers the full audio stack: text-to-speech, live voice conversations, music, sound effects, voice changing and transcription. Audio runs on ElevenLabs, Sonauto and OpenAI models, all billed in credits.
Text-to-speech (ElevenLabs)
/tts turns text into natural-sounding speech with ElevenLabs Turbo v2.5.
- 23 preset voices — Aria, Sarah, George, Charlotte, River, Rachel and more — plus any custom ElevenLabs voice ID.
- 32 languages with auto-detection and a playback-speed control (0.7x–1.2x).
- Up to 40,000 characters per request (roughly 40 minutes of audio). Long texts render in the background and drop into chat when ready.
- Preview a voice before you spend credits.
Cost: 10 credits per 1,000 characters, 10-credit minimum. See /tts.
Voice Mode (real-time conversation)
Voice Mode is a hands-free spoken conversation with Krater. It answers out loud and can use the same tools as chat while you talk.
- 6 voices: Ash (default), Echo, Verse, Alloy, Coral and Shimmer.
- 0.25 credits per second — about 15 credits per minute — billed on actual talk time. Connection setup is free.
- Sessions run up to 30 minutes. Credits are reserved up front and the unused part is refunded when you hang up.
- Requires a paid plan.
See Voice Mode.
Music (Sonauto)
/sound in Music mode generates complete songs with Sonauto V3.
- Full tracks with vocals and auto-written lyrics — or supply your own.
- Instrumental toggle for tracks without vocals.
- Describe genre, mood and style in the prompt.
- Playback starts while the song is still generating.
Cost: 15 credits per song. See /sound.
Sound effects (ElevenLabs)
/sound in Effect mode creates any sound effect from a text description with ElevenLabs Sound Effects v2.
- Clips of 1–30 seconds (5-second default).
- Priced by duration: about 1 credit per 2 seconds, 2 credits minimum.
Use cases: video production, game development, podcasts, presentations.
Voice changer (ElevenLabs)
/voice re-voices an uploaded recording with ElevenLabs Voice Changer — your performance, a different voice.
- The same 23 preset voices as
/tts, or a custom ElevenLabs voice ID. - Optional background-noise removal.
- Uploads up to 10 minutes.
Cost: 60 credits per minute of output audio, 5 credits minimum. See /voice.
Transcription
/transcribe turns audio or video files into timestamped text (Whisper V3).
- Files up to 4 hours.
- 2 credits per minute, 3-credit minimum (covers the first 90 seconds).
- The dictation mic in chat (ElevenLabs Scribe) bills 1 credit per minute, 3-credit minimum.
See /transcribe.
Cost overview
| Feature | Command | Cost |
|---|---|---|
| Text-to-speech | /tts |
10 credits per 1,000 chars |
| Voice Mode | Mic button | 15 credits per minute |
| Music | /sound |
15 credits per song |
| Sound effects | /sound Effect |
~1 credit per 2 seconds, min 2 |
| Voice changer | /voice |
60 credits per minute, min 5 |
| Transcription | /transcribe |
2 credits per minute, min 3 |
Save any audio output to Keep for 1 credit, or explore the other model families in AI models.