Use OpenAI Whisper through Speechall’s unified speech-to-text API. Rates represent USD credit usage; review purchase fees in your billing quote. Check the features and limits below before switching models.
Use the hosted OpenAI whisper-1 model for file transcription without downloading model weights or managing a GPU. Compare its credit cost and other Whisper hosts below, then test a recording before choosing a model.
API example and OpenAI retirement notice reviewed November 14, 2026. Prices and capabilities below come from the Speechall catalog; check availability before sending requests.
See our same-audio Whisper test on OpenAI, Groq and Cloudflare for measured word errors and request turnaround across six hosted variants.
OpenAI Whisper costs $0.006 per minute, or $0.36 per audio hour, in Speechall credit usage. Compare it with OpenAI’s own rate on its official pricing page.
| Audio | Cost at this rate |
|---|---|
| 1 minute | $0.006 |
| 30 minutes | $0.18 |
| 1 hour | $0.36 |
| 100 hours | $36.00 |
These examples multiply audio duration by the catalog rate. They show credits consumed, not your checkout total: Speechall adds a purchase fee when you buy credits. Review the current billing quote before budgeting.
OpenAI has announced that whisper-1 will shut down on February 26, 2027 and recommends GPT-Transcribe or GPT Live Transcribe as replacements (see OpenAI’s deprecations page). Check this page for the current catalog before you build on it.
Create a Speechall API key, add credits, and set SPEECHALL_API_KEY in your server environment. Save a short English recording as audio.wav, then run this request. Use the MIME type that matches your file.
curl --fail-with-body --request POST \
"https://api.speechall.com/v1/transcribe?model=openai.whisper-1&language=en&output_format=text" \
--header "Authorization: Bearer $SPEECHALL_API_KEY" \
--header "Content-Type: audio/wav" \
--data-binary @audio.wavA successful request returns the transcript as plain text. The model ID is openai.whisper-1, including the provider prefix. For another input language, replace en with its language code; use auto only when the catalog lists automatic language detection. Keep your API key on your server.
Already using an OpenAI audio SDK? Follow the OpenAI-compatible upload guide to change the base URL, key and model ID. The API reference lists the supported parameters.
General-purpose speech recognition model. Based on the open-source Whisper large-v2 model, offering faster performance than the open-source version.
openai.whisper-1
Use this ID when making API calls to reference this model
OpenAI
general
premium
March 1, 2023
$0.36000/audio hour
$0.00010/audio second
25.00 MB
Supported capabilities and functionalities
Input/output formats and technical details
These hosts offer different models from the Whisper family, not identical deployments: whisper-1, Large v3 and Large v3 Turbo should be evaluated separately. Each catalog entry has its own rate, file limit and subtitle support. Rates below are Speechall credit usage, not a quote for buying directly from the host.
| Model | Host | Per hour | Per minute | File limit | Subtitles |
|---|---|---|---|---|---|
| Groq Whisper Large v3 | Groq | $0.111 | $0.00185 | 25 MB | None |
| Groq Whisper Large v3 Turbo | Groq | $0.04 | $0.00067 | 25 MB | None |
| Cloudflare Whisper | Cloudflare | $0.027 | $0.00045 | 25 MB | VTT |
| Cloudflare Whisper Large v3 Turbo | Cloudflare | $0.0306 | $0.00051 | 9.5 MB | VTT |
| Cloudflare Whisper Tiny (EN) | Cloudflare | $0.027 | $0.00045 | Not listed | VTT |
| OpenAI Whisper | OpenAI | $0.36 | $0.006 | 25 MB | SRT, VTT |
Start with a short recording that represents your language, background noise and speaker overlap. Compare models on your own audio, then open the playground to try the same recording with each candidate. Playground access requires a Speechall account.
| Check | What to compare |
|---|---|
| Transcript quality | Check names, numbers, punctuation and missed speech against a transcript you have reviewed. Price alone does not establish accuracy. |
| Limits | Check each model’s file-size and duration limits in its model details. A missing limit does not mean unlimited uploads. |
| Subtitles | This model’s catalog lists: SRT, VTT. For supported models, request output_format=srt or output_format=vtt; see subtitle examples. |
| Live audio | File transcription and live streaming use different endpoints. Check the selected model’s Streaming capability and the WebSocket guide. |
| Cost | Use the same audio duration to compare credit usage, then include the purchase fee in your budget. |
To test another host through the file endpoint, replace the model query parameter with its full ID from the table. Recheck languages, output formats and optional parameters before switching production traffic; model-specific features do not automatically carry over.
GPT-Transcribe replaces GPT-4o Transcribe, and OpenAI lists whisper-1 for shutdown too. In this catalog GPT-Transcribe is $0.27 per audio hour against $0.36 for Whisper. File limits are 25 MB for GPT-Transcribe and 25 MB for Whisper, and subtitle formats are None for GPT-Transcribe and SRT, VTT for Whisper.