> ## Documentation Index > Fetch the complete documentation index at: https://www.edenai.co/docs/llms.txt > Use this file to discover all available pages before exploring further. # OpenAI-Compatible Speech to Text API > Convert speech to text with Whisper, Deepgram Nova and Mistral Voxtral through Eden AI's OpenAI-compatible /v3/audio/transcriptions endpoint. export const TechArticleSchema = ({title, description, path, articleSection, about, proficiencyLevel = "Beginner", dependencies, keywords = [], datePublished, dateModified, image, inLanguage = "en"}) => { const baseUrl = "https://www.edenai.co/docs"; const canonicalUrl = `${baseUrl}/${path}`.replace(/\/+$/, ""); const ogParams = new URLSearchParams({ division: articleSection || "", title: title || "", description: description || "" }); const resolvedImage = image || `https://edenai.mintlify.app/_mintlify/api/og?${ogParams.toString()}`; const data = { "@context": "https://schema.org", "@type": "TechArticle", "@id": `${canonicalUrl}#techarticle`, mainEntityOfPage: { "@type": "WebPage", "@id": canonicalUrl }, headline: title, name: title, description: description, url: canonicalUrl, inLanguage: inLanguage, isPartOf: { "@type": "WebSite", name: "Eden AI Documentation", url: baseUrl }, author: [{ "@type": "Organization", name: "Eden AI", url: "https://www.edenai.co/" }], publisher: { "@type": "Organization", name: "Eden AI", url: "https://www.edenai.co/", logo: { "@type": "ImageObject", url: "https://www.edenai.co/assets/logo.png" } } }; if (articleSection) data.articleSection = articleSection; if (about) data.about = { "@type": "Thing", name: about }; if (proficiencyLevel) data.proficiencyLevel = proficiencyLevel; if (dependencies) data.dependencies = dependencies; if (keywords && keywords.length) data.keywords = keywords; if (datePublished) data.datePublished = datePublished; if (dateModified) data.dateModified = dateModified; data.image = Array.isArray(resolvedImage) ? resolvedImage : [resolvedImage]; const json = JSON.stringify(data); const schemaId = `techarticle-${canonicalUrl}`; React.useEffect(() => { if (typeof document === "undefined") return; document.querySelectorAll(`script[data-schema-id="${schemaId}"]`).forEach(n => n.remove()); const script = document.createElement("script"); script.type = "application/ld+json"; script.dataset.schemaId = schemaId; script.textContent = json; document.head.appendChild(script); return () => script.remove(); }, [json, schemaId]); return null; }; Transcribe audio through Eden AI's OpenAI-compatible transcriptions endpoint. Point any OpenAI client at `https://api.edenai.run/v3` and `client.audio.transcriptions` works as a drop-in replacement for OpenAI's speech to text. ## Overview | Method | Path | Purpose | | - | - | - | | `POST` | `/v3/audio/transcriptions` | Transcribe an audio file. Returns the transcript as JSON | | `GET` | `/v3/audio/transcriptions/models` | Transcription models available in your region, as `provider/model` ids | The call is synchronous: the transcript comes back in the response, with Eden's `cost` (USD) and `provider` fields added. Transcription models are only available on `/v3/audio/transcriptions`. They are not in `GET /v3/models`, which lists chat models. This is different from [Expert Model speech to text](/docs/v3/expert-models/features/audio/speech-to-text-async), an asynchronous job API through `/v3/universal-ai/async` that covers more providers (Amazon, AssemblyAI, Gladia, Google, Microsoft and others) and options such as speaker diarization. Use `/v3/audio/transcriptions` when you want an OpenAI-compatible, drop-in transcription surface. ## Models | Provider | Model string | | - | - | | Groq | `groq/whisper-large-v3`, `groq/whisper-large-v3-turbo` | | Deepgram | `deepgram/nova-3`, `deepgram/nova-2`, `deepgram/whisper-large` | | Mistral | `mistral/voxtral-mini-latest` | Call `GET /v3/audio/transcriptions/models` for the full list available in your region, with the per-second price of each model. ## Transcribing a file The endpoint takes the audio in one of two ways, chosen by the request's `Content-Type`: * `multipart/form-data`: the audio is uploaded as a `file` part. This is what the OpenAI SDK sends. * `application/json`: the audio is referenced by `file_url` (an `https://` URL or a `data:audio/...;base64,...` URL) or by `file_id` (a file uploaded to Eden AI through [File Upload](/docs/v3/llms/file-upload)). Set exactly one of the two. ### Uploading the file The official OpenAI SDK works unchanged once `base_url` points at Eden AI: ```python OpenAI SDK theme={null} from openai import OpenAI client = OpenAI(base_url="https://api.edenai.run/v3", api_key="YOUR_API_KEY") with open("recording.wav", "rb") as audio_file: transcript = client.audio.transcriptions.create( model="groq/whisper-large-v3-turbo", file=audio_file, ) print(transcript.text) # Eden AI's extra fields are on the same object print(transcript.model_extra.get("provider"), transcript.model_extra.get("cost")) ``` ```bash cURL theme={null} curl -X POST https://api.edenai.run/v3/audio/transcriptions \ -H "Authorization: Bearer YOUR_API_KEY" \ -F model="groq/whisper-large-v3-turbo" \ -F file="@recording.wav" ``` ### Referencing a URL ```python Python theme={null} import requests url = "https://api.edenai.run/v3/audio/transcriptions" headers = { "Authorization": "Bearer YOUR_API_KEY", "Content-Type": "application/json", } payload = { "model": "groq/whisper-large-v3-turbo", "file_url": "https://raw.githubusercontent.com/openai/whisper/main/tests/jfk.flac", "language": "en", } response = requests.post(url, headers=headers, json=payload) result = response.json() print(result["text"]) print("Provider:", result["provider"], "| Cost (USD):", result.get("cost")) ``` ```bash cURL theme={null} curl -X POST https://api.edenai.run/v3/audio/transcriptions \ -H "Authorization: Bearer YOUR_API_KEY" \ -H "Content-Type: application/json" \ -d '{ "model": "groq/whisper-large-v3-turbo", "file_url": "https://raw.githubusercontent.com/openai/whisper/main/tests/jfk.flac", "language": "en" }' ``` ### Request parameters | Field | Required | Description | | - | - | - | | `model` | yes | `provider/model`, e.g. `groq/whisper-large-v3-turbo`. The model name alone (`whisper-large-v3-turbo`) also works, through [Provider Routing](/docs/v3/llms/provider-routing) | | `file` | one of three | The audio file, as a multipart upload | | `file_url` | one of three | JSON only. An `https://` URL or a `data:audio/...;base64,...` URL | | `file_id` | one of three | JSON only. The id of a file uploaded to Eden AI | | `language` | no | ISO-639-1 code of the spoken language, e.g. `en`. Omit it to let the model detect the language | | `prompt` | no | Text that guides the model's style or continues the context of a previous segment | | `response_format` | no | `json` (default) or `verbose_json`. See [Response format](#response-format) | | `timestamp_granularities` | no | `["segment"]` and/or `["word"]`, with `verbose_json`. Honored by the models that support it | | `temperature` | no | Sampling temperature between 0 and 1 | Unknown fields are ignored. The audio can be at most 25 MB. ## Response format ```json theme={null} { "text": "And so, my fellow Americans, ask not what your country can do for you, ask what you can do for your country.", "usage": null, "cost": 0.00012221, "provider": "groq" } ``` With `"response_format": "verbose_json"`, the response also carries the detected `language`, the audio `duration` in seconds, and `segments` with start and end times. Models that support word timestamps add `words` when `timestamp_granularities` includes `"word"`. Use `json` or `verbose_json`. Unlike OpenAI, `text`, `srt` and `vtt` do not return a plain-text or subtitle body. The Groq and Deepgram models ignore them and return the JSON transcript; other models may return an error. With the OpenAI SDK, keep the default `json` format: passing `response_format="text"` makes the SDK return the raw JSON string instead of a `Transcription` object. ## Listing transcription models ```bash cURL theme={null} curl https://api.edenai.run/v3/audio/transcriptions/models \ -H "Authorization: Bearer YOUR_API_KEY" ``` Each entry carries its `pricing`, e.g. `{"input_cost_per_second": 0.00001111}` for `groq/whisper-large-v3-turbo`. The endpoint takes the same `view` parameter as the other [model listings](/docs/v3/llms/listing-models). ## Pricing Transcription is billed per second of audio, at the rate `GET /v3/audio/transcriptions/models` shows: an 11-second clip on `groq/whisper-large-v3-turbo` costs 11 × $0.00001111 ≈ $0.000122. The amount charged is in the response's `cost` field. ## Limitations * **Files up to 25 MB.** A larger file returns `413`. A file whose duration cannot be read returns `400`. * **No `@edenai` routing.** `"model": "@edenai"` returns `400`; name a model. * **No `fallbacks`.** The field is ignored, like any unknown field: only `model` is tried, and a failure is not retried on another model. * **No streaming.** The transcript is returned in one response once the whole file is transcribed. * **No translations endpoint.** `/v3/audio/translations` (speech to English text) is not implemented. ## Next Steps Generate speech through the OpenAI-compatible speech endpoint Asynchronous transcription with more providers and diarization This documentation is built and hosted on [Mintlify](https://mintlify.com), a developer documentation platform.