Confronta e prova modelli open weight per testo, codice, voce e visione, senza registrazione. Poi passa in produzione con la stessa API compatibile con OpenAI.
Whisper è un modello avanzato per il riconoscimento vocale automatico (ASR) e la traduzione, proposto nel documento "Robust Speech Recognition via Large-Scale Weak Supervision" di Alec Radford et al. di OpenAI. Addestrato per oltre 5 milioni di ore con dati etichettati, Whisper dimostra una forte capacità di generalizzare su molti dataset e domini in un contesto zero-shot. Whisper large-v3-turbo è una versione ottimizzata di un Whisper large-v3 snellito. In altre parole è esattamente lo stesso modello, ma il numero di livelli di decodifica è stato ridotto da 32 a 4. Di conseguenza il modello è molto più veloce, a discapito però di una leggera perdita di qualità.
The OpenAI-compatible Audio transcription API endpoint allows you to recognize and transcribe audio, especially human speech, into text.
Audio transcription is a technology that converts spoken language into written text. It is a complex process that involves several stages, including speech signal preprocessing, feature extraction, acoustic modeling, language modeling, and speech recognition engine.
AI Endpoints makes it easy, with ready-to-use inference APIs. Discover how to use them:
The API accepts multiple audio formats, including:
Make sure your file is in one of these formats before sending it for transcription.
The Audio transcription endpoint offers you a wide range of transcription options for audios. Learn how to use them with the following examples:
First, export your access token to the OVH_AI_ENDPOINTS_ACCESS_TOKEN environment variable:
export OVH_AI_ENDPOINTS_ACCESS_TOKEN=<your-access-token>If you do not have an access token key yet, follow the instructions in the AI Endpoints – Getting Started.
The whisper-large-v3-turbo API is compatible with the OpenAI specification.
First install the openai library:
pip install openaiNext, export your access token to the OVH_AI_ENDPOINTS_ACCESS_TOKEN environment variable:
export OVH_AI_ENDPOINTS_ACCESS_TOKEN=<your-access-token>Finally, run the following Python code:
import os
from openai import OpenAI
url = "https://oai.endpoints.kepler.ai.cloud.ovh.net/v1/"
audio_file_path = "my_audio.mp3"
client = OpenAI(
base_url=url,
api_key=os.getenv('OVH_AI_ENDPOINTS_ACCESS_TOKEN')
)
audio_file = open(audio_file_path, "rb")
transcript_answer = client.audio.transcriptions.create(
model="whisper-large-v3-turbo",
file=audio_file,
temperature=0.0
)
print(transcript_answer)Which returns the following:
Transcription(
text=" It is very exciting that the Olympic Games will be held in Paris in 2024 for France [...] and to cheer on their home country.",
logprobs=None,
task='transcribe',
success=True,
language='en',
duration=60.936,
words=[],
segments=[
{
'id': 1,
'seek': 0,
'start': 0.0,
'end': 8.32,
'text': ' It is very exciting that the Olympic Games will be held in Paris in 2024 for France.',
'tokens': [50365, 467, 307, 588, 4670, 300, 264, 19169, 12761, 486, 312, 5167, 294, 8380, 294, 45237, 337, 6190, 13, 50793],
'temperature': 0.0,
'avg_logprob': -0.10994318,
'compression_ratio': 1.4497042,
'no_speech_prob': 0.0
},
...,
{
'id': 8,
'seek': 5592,
'start': 56.48,
'end': 60.16,
'text': ' cheer, cheer on their home country.',
'tokens': [50365, 12581, 11, 12581, 322, 641, 1280, 1941, 13, 50584],
'temperature': 0.0,
'avg_logprob': -0.56960225,
'compression_ratio': 0.875,
'no_speech_prob': 0.00919342
}
],
diarization=[],
usage={
'type': 'duration',
'duration': 61.0
}
)To process an audio file with curl, run the following command in your terminal:
curl -X POST "https://oai.endpoints.kepler.ai.cloud.ovh.net/v1/audio/transcriptions" \
-H "accept: application/json" \
-H "Authorization: Bearer $OVH_AI_ENDPOINTS_ACCESS_TOKEN" \
-F "model=whisper-large-v3-turbo" \
-F "file=@my_audio.mp3" \
-F "temperature=0.0" \
-H 'Content-Type: multipart/form-data'Which returns the following:
{
"task": "transcribe",
"success": true,
"language": "en",
"duration": 60.936,
"text": " It is very exciting that the Olympic Games will be held in Paris in 2024 for France [...] and to cheer on their home country.",
"words": [],
"segments": [
{
"id": 1,
"seek": 0,
"start": 0,
"end": 8.48,
"text": " It is very exciting that the Olympic Games will be held in Paris in 2024 for France.",
"tokens": [50365,467,307,588,4670,300,264,19169,12761,486,312,5167,294,8380,294,45237,337,6190,13,50865],
"temperature": 0,
"avg_logprob": -0.31054688,
"compression_ratio": 0.9767442,
"no_speech_prob": 0
},
...,
{
"id": 5,
"seek": 3484,
"start": 46.12,
"end": 60.24,
"text": " It is a chance for everyone to get involved and become interested about different types of sports and to cheer on their home country.",
"tokens": [50885,467,307,257,2931,337,1518,281,483,3288,293,1813,3102,466,819,3467,295,6573,293,281,12581,322,641,1280,1941,13,51640],
"temperature": 0,
"avg_logprob": -0.11174939,
"compression_ratio": 1.4268292,
"no_speech_prob": 0
}
],
"diarization": [],
"usage": {
"type": "duration",
"duration": 61
}
}First install the requests library:
pip install requestsNext, export your access token to the OVH_AI_ENDPOINTS_ACCESS_TOKEN environment variable:
export OVH_AI_ENDPOINTS_ACCESS_TOKEN=<your-access-token>Finally, run the following Python code:
import os
import requests
url = "https://oai.endpoints.kepler.ai.cloud.ovh.net/v1/audio/transcriptions"
audio_file_path = "my_audio.mp3"
headers = {
"accept": "application/json",
"Authorization": f"Bearer {os.getenv('OVH_AI_ENDPOINTS_ACCESS_TOKEN')}",
}
files = {"file": open(audio_file_path, "rb")}
data = {
"model": "whisper-large-v3-turbo",
"temperature": 0.0,
"timestamp_granularities": "segment, word",
"response_format": "verbose_json"
}
response = requests.post(url, headers=headers, files=files, data=data)
# Handle response
if response.status_code == 200:
print(response.json())
else:
print("Error:", response.status_code, response.text)Returning the following result, which contains the audio transcription API response:
{
'task': 'transcribe',
'success': True,
'language': 'en',
'duration': 60.936,
'text': " It is very exciting that the Olympic Games will be held in Paris in 2024 for France [...] and to cheer on their home country.",
'words': [
{'word': 'It', 'start': 0.0, 'end': 0.7},
{'word': 'is', 'start': 0.7, 'end': 0.92},
{'word': 'very', 'start': 0.92, 'end': 1.18},
...,
{'word': 'their', 'start': 58.66, 'end': 59.24},
{'word': 'home', 'start': 59.24, 'end': 59.62},
{'word': 'country.', 'start': 59.62, 'end': 60.16}
],
'segments': [
{
'id': 1,
'seek': 0,
'start': 0.0,
'end': 8.32,
'text': ' It is very exciting that the Olympic Games will be held in Paris in 2024 for France.',
'tokens': [50365, 467, 307, 588, 4670, 300, 264, 19169, 12761, 486, 312, 5167, 294, 8380, 294, 45237, 337, 6190, 13, 50793],
'temperature': 0.0,
'avg_logprob': -0.10994318,
'compression_ratio': 1.4497042,
'no_speech_prob': 0.0
},
...,
{
'id': 8,
'seek': 5592,
'start': 56.48,
'end': 60.16,
'text': ' cheer, cheer on their home country.', 'tokens': [50365, 12581, 11, 12581, 322, 641, 1280, 1941, 13, 50584],
'temperature': 0.0,
'avg_logprob': -0.56960225,
'compression_ratio': 0.875,
'no_speech_prob': 0.00919342
}
],
'diarization': [],
'usage': {
'type': 'duration',
'duration': 61.0
}
}When using AI Endpoints, the following rate limits apply:
On top of the request rate limitations, AI Endpoints also applies the following limitations to the audio data:
If you exceed these limits, different HTTP error codes will be returned depending on the type of limitation exceeded:
Note that exceeding either the rate limit or audio size/duration limits will result in an appropriate error response to help you understand and address the issue.
If you require higher usage, please get in touch with us to discuss increasing your rate limits.
Want to explore the full capabilities of the ASR API? Dive into our dedicated Speech to Text guide to discover advanced parameters (such as diarization, timestamp granularities, chunking strategy, and prompts) and ready to run code examples in Python, cURL, and JavaScript.
For a broader overview of AI Endpoints, explore the full AI Endpoints Documentation.
Reach out to our support team or join the OVHcloud Discord #ai-endpoints channel to share your questions, feedback, and suggestions for improving the service, to the team and the community.
For more information about the Whisper model features, please refer to Hugging Face documentation:
New to AI Endpoints? This guide walks you through everything you need to get an access token, call AI models, and integrate AI APIs into your apps with ease.
Inizia il tutorialExplore what AI Endpoints can do. This guide breaks down current features, future roadmap items, and the platform's core capabilities so you know exactly what to expect.
Inizia il tutorialRunning into issues? This guide helps you solve common problems on AI Endpoints, from error codes to unexpected responses. Get quick answers, clear fixes, and helpful tips to keep your…
Inizia il tutorialLearn how to use OVHcloud AI Endpoints Virtual Models.
Inizia il tutorialExplore the full potential of the ASR API through advanced parameters (diarization, timestamp granularities, chunking strategies, prompts ...) and ready-to-run code examples in Python, cURL, and JavaScript.
Inizia il tutorialTurn hours of audio into sharp, readable summaries! This guide walks you through building an AI-powered audio assistant using ASR and LLMs, perfect for meetings, podcasts, or any voice recordings.
Inizia il tutorialBuild a voice-enabled assistant in under 100 lines of code! Learn how to combine ASR, LLM, and TTS endpoints to create an AI that listens, understands, and responds.
Inizia il tutorialDiscover how to synthesize realistic dialogues and automatically identify who’s speaking. This notebook shows you how to generate speech and label each speaker, perfect for building smart transcripts.
Inizia il tutorial