Speech is Cheap is an alternative to the hosted OpenAI Whisper API for long recordings and recurring transcription. Its $20 plan includes 21,600 base minutes, with URLs, uploads and optional word timing, speakers and sound labels in one job workflow. Compare it when you need more than a transcript string or are planning beyond the Whisper API sunset.
21,600 Audio Minutes per Month
$20
Speech is Cheap subscription, base transcription
Whisper and the listed GPT-4o models are scheduled to sunset on February 26, 2027. OpenAI recommends GPT-Transcribe for recorded audio.
OpenAI Model Deprecations ; OpenAI File Transcription GuideChoose the cost and workflow that fit your product.
Send a recording URL or upload a file under 2 GiB to Speech is Cheap, with up to 1,440 minutes per job. Hosted Whisper requires a file upload with a limit of about 0.024 GiB, so larger recordings need preparation and splitting.
Speech is Cheap Job Requests ; Speech is Cheap Product Facts ; OpenAI Audio API FAQ ; OpenAI File Transcription GuideEnable Parse Words and Parse Speakers on the same Speech is Cheap job. Read both under output.segments with timestamps in seconds, instead of selecting between the different OpenAI models documented for these outputs.
Speech is Cheap Job Requests ; Speech is Cheap Job Responses ; OpenAI File Transcription GuideEnable Label Audio to classify each segment, including music and silence. Use those labels in JSON or add non-speech cues to SRT and VTT captions from the same job.
Speech is Cheap Job Requests ; Speech is Cheap Job Responses ; Speech is Cheap PricingThe hosted Whisper translations endpoint can turn supported-language audio into English text. Speech is Cheap transcribes the spoken language and does not offer that translation endpoint. Account for the scheduled Whisper sunset in a new integration.
OpenAI File Transcription Guide ; OpenAI Model Deprecations ; Speech is Cheap Product FactsOpenAI offers a separate Realtime transcription path for live input. GPT-Transcribe also streams partial text from completed recordings. Use its live products when microphone input is essential; Speech is Cheap processes existing recordings.
OpenAI Realtime Transcription ; OpenAI File Transcription Guide ; Speech is Cheap Job RequestsOpen-source Whisper can run in your own environment under its published license. Choose that path when controlling deployment outweighs operating your own transcription infrastructure. Those costs are not the hosted API prices shown here.
OpenAI Open-Source WhisperCompare base transcription and the add-ons your product needs.
| Audio Volume | Speech is Cheap PAYG | Speech is Cheap Subscription | OpenAI whisper-1 (Sunset) | OpenAI GPT-Transcribe | OpenAI 4o Mini (Estimate, Sunset) | OpenAI 4o Diarize (Estimate, Sunset) |
|---|---|---|---|---|---|---|
| 12,500 minutes | $25.00 | $20.00 | $75.00 | $56.25 | $37.50 | $75.00 |
| 25,000 minutes | $50.00 | $23.15 | $150.00 | $112.50 | $75.00 | $150.00 |
| 50,000 minutes | $100.00 | $46.30 | $300.00 | $225.00 | $150.00 | $300.00 |
| 100,000 minutes | $200.00 | $92.60 | $600.00 | $450.00 | $300.00 | $600.00 |
| Audio and Features | Speech is Cheap PAYG | Speech is Cheap Subscription | OpenAI whisper-1 (Sunset) | OpenAI GPT-Transcribe | OpenAI 4o Mini (Estimate, Sunset) | OpenAI 4o Diarize (Estimate, Sunset) |
|---|---|---|---|---|---|---|
| 12,500 minutes + words | $37.50 | $26.25 | $75.00 | -- | -- | -- |
| 12,500 minutes + speakers | $50.00 | $32.50 | -- | -- | -- | $75.00 |
| 25,000 minutes + words | $75.00 | $35.65 | $150.00 | -- | -- | -- |
| 25,000 minutes + speakers | $100.00 | $48.15 | -- | -- | -- | $150.00 |
| 50,000 minutes + words | $150.00 | $71.30 | $300.00 | -- | -- | -- |
| 50,000 minutes + speakers | $200.00 | $96.30 | -- | -- | -- | $300.00 |
| 100,000 minutes + words | $300.00 | $142.60 | $600.00 | -- | -- | -- |
| 100,000 minutes + speakers | $400.00 | $192.60 | -- | -- | -- | $600.00 |
Choose your volume and add-ons. Each step doubles the audio minutes.
$92.60
per month
$200.00
per month
$600.00
per month
$450.00
per month
$300.00
per month
$600.00
per month
-- means no verified price for the selected feature combination in the tables and calculator.
Put long recordings, sound labels and progress updates to work in your app.
| Compare | Speech is Cheap | OpenAI Whisper API |
|---|---|---|
| Audio Input | Recorded audio and video. Progress SSE returns results for an existing recording. Speech is Cheap Job Requests | File transcription and a separate Realtime API. Open-source Whisper is a model you deploy yourself, not the hosted service. OpenAI File Transcription Guide ; OpenAI Realtime Transcription ; OpenAI Open-Source Whisper |
| File Limits | 0.1 to 1,440 minutes per recording. Direct uploads: under 2 GiB. Speech is Cheap Product Facts | Whisper uploads are about 0.024 GiB per request. The file API requires uploaded audio rather than an input URL. Check model-specific limits for newer routes. OpenAI Audio API FAQ ; OpenAI File Transcription Guide |
| Languages | 100 supported languages; optional detection per speech segment. Speech is Cheap Product Facts ; Speech is Cheap Job Requests | Multilingual transcription; support differs by model. Whisper also has an endpoint that translates supported-language audio into English. OpenAI File Transcription Guide ; OpenAI Whisper Model and Price |
| Outputs | JSON, SRT and VTT. Word timestamps and voice-based speaker labels are paid add-ons. Speech is Cheap Job Responses ; Speech is Cheap Pricing | Whisper: JSON, text, SRT, VTT and verbose JSON with word timing. Diarization uses a separate model and diarized JSON. Newer models have different format support. OpenAI File Transcription Guide |
| Integration | URL or multipart upload, REST, polling, completion webhooks and progress SSE. Speech is Cheap Job Requests ; Speech is Cheap Job Responses | Multipart file submission through REST or SDKs. Read the response; selected newer models can stream text events. OpenAI File Transcription Guide |
| Audio Labels | Label Audio adds a structured label to each segment, including music and silence. Non-speech sounds appear as caption cues in SRT and VTT. Speech is Cheap Job Requests ; Speech is Cheap Pricing | An equivalent structured sound label on each transcript segment is not verified for the transcription routes priced here. OpenAI File Transcription Guide |
| Progress While Processing | Progress updates and chronological transcript segments over SSE from the same recorded job. Speech is Cheap Job Requests | Whisper does not support streaming. GPT-Transcribe and the diarization model can stream file-transcription events; live audio uses Realtime. OpenAI File Transcription Guide ; OpenAI Audio API FAQ ; OpenAI Realtime Transcription |
| Support | Initial email response within one business day for subscribers, three for pay as you go. Resolution time is not guaranteed. Speech is Cheap Service Commitments | Public API documentation and account help. Model lifecycle notices identify replacement paths and shutdown dates. OpenAI Audio API FAQ ; OpenAI Model Deprecations |
Use public timing data to plan your recording workflow.
Speech is Cheap's public Benchmarks lets you inspect processing times by recording length and selected add-ons, with sample counts. Use those cohorts to evaluate the kind of audio your product processes.
Replace the uploaded-file response with a recorded job that can accept a URL. Use the mapping to keep subtitles, word timing and speaker output connected to your application.
| Compare | OpenAI Whisper API | Speech is Cheap |
|---|---|---|
| Submit | POST /v1/audio/transcriptions Multipart: file, model Authorization: Bearer KEYOpenAI File Transcription Guide | POST <api|upload>/v2/jobs/ api: JSON input_url upload: multipart input_file Authorization: Bearer KEYSpeech is Cheap Job Requests |
| Speaker Labels | gpt-4o-transcribe-diarize response_format: diarized_json segments[].speakerOpenAI File Transcription Guide | can_parse_speakers: true output.segments[].speaker_idSpeech is Cheap Job Requests ; Speech is Cheap Job Responses |
| Word Timing | whisper-1; verbose_json timestamp_granularities[]: word words[].start / end in secondsOpenAI File Transcription Guide | can_parse_words: true output.segments[].words[] start / end in secondsSpeech is Cheap Job Requests ; Speech is Cheap Job Responses |
| Status and Text | Whisper: response text / segments Newer models: JSON or eventsOpenAI File Transcription Guide | PENDING / COMPLETED / FAILED CANCELED / UNKNOWN output.segments[].textSpeech is Cheap Job Responses |
| Completion | Wait for the file response Or stream=true on supported models Whisper: no streamingOpenAI File Transcription Guide ; OpenAI Audio API FAQ | Webhook: job output Poll GET /v2/jobs/:id Or monitor progress SSE: can_stream_output: trueSpeech is Cheap Job Requests ; Speech is Cheap Job Responses |
api means https://api.speechischeap.com with JSON input_url. upload means https://upload.speechischeap.com with multipart input_file.
A successful upload always returns an SSE stream with text/event-stream, even with can_stream_output: false. In that mode, the final event contains the asynchronous job response.
curl --request POST \
--url https://api.speechischeap.com/v2/jobs/ \
--header 'Authorization: Bearer KEY' \
--header 'Content-Type: application/json' \
--data '{
"input_url": "https://example.com/recording.mp3",
"webhook_url": "https://your-domain.com/transcription-webhook",
"can_parse_speakers": true,
"can_parse_words": true
}'Example request. Replace the URLs and API key. Both add-ons are billed.
Submit input_url when your recording is already hosted. For direct upload, use the Speech is Cheap multipart interface. Receive a job ID and handle PENDING, COMPLETED and failure states rather than assuming the first response contains the full transcript.
Speech is Cheap Job Requests ; Speech is Cheap Job Responses ; OpenAI File Transcription GuideMap verbose Whisper words and segments into output.segments. Enable both word timing and speakers when your application needs them. Set output_format for native SRT or VTT; a different parameter name does not imply identical caption segmentation.
OpenAI File Transcription Guide ; Speech is Cheap Job Requests ; Speech is Cheap Job ResponsesCheck language selection, prompts and English translation separately when mapping the inputs your application sends.
OpenAI File Transcription Guide ; OpenAI Model Deprecations ; OpenAI GPT-Transcribe ModelReview retention and private mode when saving transcripts. Speech is Cheap results are retrievable by job ID, so treat those IDs as sensitive.
A Production Migration
Metacast
Metacast's result versus its previous, unnamed provider
Podcast Transcription
When Metacast's previous provider stopped completing transcription jobs, its team moved production to Speech is Cheap over a weekend. The new API became its primary transcription engine.
Read the Case Study"It's now our primary transcription engine."
Speech is Cheap is one option for recorded files, with a job workflow, long-file support, captions and optional word timing and speakers. OpenAI recommends GPT-Transcribe for its own recorded-audio path. Choose based on the exact outputs your application needs.
Yes. Enable Parse Speakers and Parse Words in Speech is Cheap. Both are paid add-ons, and the resulting segments and words use seconds for JSON timestamps. The calculator includes both charges.
The prices compare hosted APIs. Open-source Whisper is separately available to run yourself. It has no hosted API fee, but you supply the infrastructure and handle operations, capacity and output processing.
OpenAI schedules whisper-1 and the GPT-4o transcription models for removal on February 26, 2027. Inventory the formats and timestamps your application consumes, evaluate representative recordings, and update the parser and delivery workflow before routing production jobs.
The linked sources support the prices, product boundaries and migration details on this page.
OpenAI documents streamed text events for newer file-transcription models, while Whisper waits for the result. Receiving the first text and completing the entire recording are different measurements.
Speech is Cheap Benchmarks ; OpenAI File Transcription GuideChecked
Checked
Checked
Checked
Checked
Checked
Checked
Checked
Checked
Checked
Checked
Checked
Checked
Checked
Checked
Select Subscription or PAYG on the pricing page. Check included minutes, overage and add-ons for the workload you compared above.
All prices are in USD. Examples use single-channel, whole-minute files no longer than 10 minutes, with selected features on every file. Audio must meet each model's format and size requirements. These are calculated prices, not a claim of equal accuracy or throughput. Totals are rounded to cents.
Speech is Cheap PAYG costs $0.002 per billed minute. Subscription costs $20 per month for 21,600 base minutes, then $0.000926 per extra minute. Unused included minutes expire each month.
Speech is Cheap rounds every file up to a whole minute for base transcription and all enabled add-ons. Subscription totals include the monthly fee, base overage and add-ons on all enabled-job minutes. Partial-minute files increase its billed minutes.
Parse Speakers adds $0.002 per PAYG minute or $0.001 on subscription. Word-Level Timestamps adds $0.001 per PAYG minute or $0.0005 on subscription. Label Audio adds $0.0002 per PAYG minute or $0.0001 on subscription.
Speech is Cheap PricingHosted Whisper costs $0.006 per minute and GPT-Transcribe costs $0.0045. GPT-4o Mini and GPT-4o Diarize columns use minute-based estimates of $0.003 and $0.006 for token-priced models; actual bills can vary. These GPT-4o models and whisper-1 are scheduled to sunset on February 26, 2027.
OpenAI Whisper Model and Price ; OpenAI API Pricing ; OpenAI GPT-Transcribe Model ; OpenAI Diarization Model ; OpenAI Model DeprecationsThe listed OpenAI models do not have a verified single-call configuration that returns both word timestamps and speaker annotations. Separate inference passes, external diarization, translation and self-hosting are outside these estimates.
OpenAI File Transcription Guide ; OpenAI Open-Source WhisperThe Audio API FAQ specifies 25 MiB for legacy Whisper uploads, while the file guide uses 25 MB. Both are approximately 0.024 GiB at the precision shown. This limit applies to Whisper, not every OpenAI audio endpoint.
OpenAI Audio API FAQ ; OpenAI File Transcription GuideAn equivalent per-segment sound-label output is unverified for the listed OpenAI models.
Estimates exclude tax, trial credits, negotiated discounts, multichannel charges, storage and network charges, and unselected features. Published automatic volume discounts and explicitly labeled promotions are included where stated. Paid Speech is Cheap API access requires a payment method; the free browser demo requires neither an account nor a card.
Speech is Cheap Pricing See All Speech is Cheap Prices