Skip to main content

Marketplace · 30+ models

Speech & Audio Voice in, voice out

Automatic speech recognition, text-to-speech, speaker diarization, audio classification, and music generation — for contact centers, media, and voice products.

Editor's pick

The speech & audio model most teams reach for first.

All categories
Editor's pick

Whisper Large v3

OpenAI·1.55B params·MIT

Multilingual ASR + translation across 99 languages; the open ASR default.

Deploy on YobibyteCompare on InferenceBench

Spec sheet

Family
OpenAI
Parameters
1.55B
License
MIT
Status
Live
Best for
Voice in, voice out
Sits in
Speech & Audio

Pricing and routing rank visible on InferenceBench. Variants and quantisations appear in the Yobibyte deploy console.

The rest of the lineup

5 more in Speech & Audio. All deployable in one click.

#
Model
Family
Params
License
Deploy
02
Whisper Turbo

8× faster than Large v3 with only minor accuracy loss; great for streaming.

OpenAI
809M
MIT
Deploy
03
Coqui XTTS v2

Multilingual zero-shot voice cloning TTS from a 6-second reference sample.

Coqui
467M
Coqui Public Model
Deploy
04
Pyannote 3.1

Speaker diarization pipeline: who spoke when, end-to-end.

Pyannote
—
MIT
Deploy
05
SeamlessM4T v2

Speech-to-speech, speech-to-text, and text-to-speech across 100 languages.

Meta
2.3B
CC-BY-NC 4.0
Deploy
06
MMS

Massively multilingual speech recognition covering 1,100+ languages.

Meta
1B
CC-BY-NC 4.0
Deploy

Showing 6 of 30+. The full catalog (with quantisations, hardware variants, and per-region pricing) lives in the Yobibyte console.

Quick start

Five lines to your first speech & audio call.

Every model in this category is reachable from the same Yobitel SDK. Swap the model name; the rest of the call shape stays identical. Authenticated via your workspace key.

Get an API keySDK on GitHub
speech-audio-quickstart.py
PythonTypeScriptcURL
from yobitel import Inference

# Whisper Large v3 — multilingual ASR
client = Inference(model="openai/whisper-large-v3")

result = client.transcribe(
    audio="board_meeting_2025_03_14.wav",
    language="en",
    diarise=True,           # who-spoke-when
    word_timestamps=True,
)

for seg in result.segments:
    print(f"[{seg.speaker}] {seg.text}")

Where teams ship this

Real speech & audio. In production.

Four use cases that customers run today. Pick a model from the lineup above, deploy on Yobibyte, plug it into the surrounding stack. Done.

  1. 01

    Meeting and call transcription

  2. 02

    Voice assistants and IVR

  3. 03

    Contact-center analytics and QA

  4. 04

    Content dubbing and accessibility

Frameworks

Bring what your team already knows

faster-whisperCoqui TTSPyannoteTriton

Yobitel handles the serving layer (GPU scheduling, KV cache, autoscaling, request batching) so your team focuses on the model and the product.

Learn about Yobibyte

Explore the rest

Other categories in the marketplace

85+

Computer Vision

Object detection, image classification, segmentation

120+

NLP & Language

Text generation, translation, sentiment, summarization

95+

Generative AI

Image gen, text gen, code gen, multimodal

60+

Data Analytics

Predictive analytics, forecasting, anomaly detection

45+

Automation & RPA

Process automation, workflow AI, document processing

50+

Industry-Specific

Vertical-specific models by industry

25+

Recommendation

Recommender systems, personalization, content matching

Don't see what you need?

Bring your own model or fine-tune one of ours. Yobitel engineers can sit with your team and ship the right stack.

Start BuildingContact Sales