Microsoft AI MAI-Transcribe 1.5 is a speech-to-text model for converting recorded audio into written content. Its exact model card describes recognition across 43 languages, automatic language identification and robustness across accents, dialects, speaking styles and noisy conditions. The edition expands the 25-language set of MAI-Transcribe 1 by eighteen languages. A useful result is the written speech from a lecture, an interview, a call or a voice note, ready to check against the recording before a later editing or analysis step.
Prepare a complete, intelligible recording in one of the maker’s listed formats: WAV, MP3 or FLAC. Keep the original file and any division into sections so the transcript can be compared with the audio. For a lecture, retain the phrases around a technical term rather than extracting only the difficult word. For a voice note, preserve a correction that changes a name, time or number. The surrounding speech provides context for recognizing what was said, and the listener’s review can identify where the output needs correction.
Automatic language identification lets a recording be processed without requiring a language choice in advance. The 43-language set includes English, French, German, Italian, Spanish, Hindi, Chinese, Japanese, Korean, Arabic, Indonesian, Russian, Thai and Vietnamese. The added languages include Bulgarian, Catalan, Greek, Ukrainian and several South Asian languages. Check that the intended language belongs to this edition’s set rather than borrowing the broader coverage of MAI-Transcribe 2. For multilingual collections, retain the recording’s language with the reviewed text so downstream editing follows the actual spoken content.
The 1.5 model card describes keyword or entity biasing with up to 200 terms, and its release explains the purpose: make likely names, acronyms and specialist words available as context. The maker describes contextual hints rather than blindly forcing supplied words into the result. Prepare a relevant vocabulary, such as the project name, a technical instrument or the correct spelling of a person discussed in the recording. Use that terminology as context for the actual recording, and check the recognized terms against the speech.
For a recorded engineering lesson, prepare the names of the instruments discussed and the exact spelling of any project acronym. Write their exact spellings in a focused terminology list and keep the list alongside the audio and transcript. Review those occurrences against the spoken passages rather than treating a vocabulary entry as proof of a match. Names that sound similar can still need checking. The result should preserve the words actually spoken, including a correction or qualification that changes their meaning.
The manufacturer emphasizes performance on real-world audio, including background noise and diverse microphone conditions. Practical review can focus on passages with simultaneous speech, faint voices or words partly covered by another sound. Keep uncertainties tied to the corresponding audio passage instead of filling them from assumptions about the topic. A plausible sentence may still contain the wrong proper noun or number. For a call transcript, the useful deliverable is a checked record of the spoken content that a later analysis can use, rather than a guessed account of the caller’s intent.
The exact 1.5 model card lists a maximum file size of 300 MB and two hours. Its linked synchronous transcription REST reference describes a stricter serving input: smaller than 250 MB and shorter than two hours. These are separately stated model-card and interface limits. Prepare files within the allowance of the intended interface and organize longer material at meaningful recording boundaries. A file-size limit concerns the encoded audio, while the duration limit concerns the speech recording’s length; neither defines a text-model context window.
The model card’s sample output is JSON containing the recording duration and combined recognized text. The text can become an accessible transcript, a script for captions or an input for later summarization and search. Caption cue grouping and alignment to the recording are separate preparation when the result does not include the required timing data. For interview records, attributing a line to a named participant needs a separate checked record of who spoke. Preserve the recognized wording and the audio as the basis for those later artifacts.
The edition’s boundaries matter when planning the output. The 1.5 card explicitly places speaker diarization outside its supported capabilities, and the maker’s version comparison assigns word-level timestamps and clean/verbatim style controls to MAI-Transcribe 2. A native streaming model is also a separate edition. This 1.5 profile therefore centres on recognized text from recordings and its documented domain context, without turning another edition’s speaker, timing or live-input features into part of the result. Review names, numbers, technical terms and unclear speech before preparing a quoted passage, a caption track or a derived document.
", "aboutText": "Microsoft AI MAI-Transcribe 1.5 is a speech-to-text model for converting recorded audio into written content. Its exact model card describes recognition across 43 languages, automatic language identification and robustness across accents, dialects, speaking styles and noisy conditions. The edition expands the 25-language set of MAI-Transcribe 1 by eighteen languages. A useful result is the written speech from a lecture, an interview, a call or a voice note, ready to check against the recording before a later editing or analysis step.\n\nPrepare a complete, intelligible recording in one of the maker’s listed formats: WAV, MP3 or FLAC. Keep the original file and any division into sections so the transcript can be compared with the audio. For a lecture, retain the phrases around a technical term rather than extracting only the difficult word. For a voice note, preserve a correction that changes a name, time or number. The surrounding speech provides context for recognizing what was said, and the listener’s review can identify where the output needs correction.\n\nAutomatic language identification lets a recording be processed without requiring a language choice in advance. The 43-language set includes English, French, German, Italian, Spanish, Hindi, Chinese, Japanese, Korean, Arabic, Indonesian, Russian, Thai and Vietnamese. The added languages include Bulgarian, Catalan, Greek, Ukrainian and several South Asian languages. Check that the intended language belongs to this edition’s set rather than borrowing the broader coverage of MAI-Transcribe 2. For multilingual collections, retain the recording’s language with the reviewed text so downstream editing follows the actual spoken content.\n\nThe 1.5 model card describes keyword or entity biasing with up to 200 terms, and its release explains the purpose: make likely names, acronyms and specialist words available as context. The maker describes contextual hints rather than blindly forcing supplied words into the result. Prepare a relevant vocabulary, such as the project name, a technical instrument or the correct spelling of a person discussed in the recording. Use that terminology as context for the actual recording, and check the recognized terms against the speech.\n\nFor a recorded engineering lesson, prepare the names of the instruments discussed and the exact spelling of any project acronym. Write their exact spellings in a focused terminology list and keep the list alongside the audio and transcript. Review those occurrences against the spoken passages rather than treating a vocabulary entry as proof of a match. Names that sound similar can still need checking. The result should preserve the words actually spoken, including a correction or qualification that changes their meaning.\n\nThe manufacturer emphasizes performance on real-world audio, including background noise and diverse microphone conditions. Practical review can focus on passages with simultaneous speech, faint voices or words partly covered by another sound. Keep uncertainties tied to the corresponding audio passage instead of filling them from assumptions about the topic. A plausible sentence may still contain the wrong proper noun or number. For a call transcript, the useful deliverable is a checked record of the spoken content that a later analysis can use, rather than a guessed account of the caller’s intent.\n\nThe exact 1.5 model card lists a maximum file size of 300 MB and two hours. Its linked synchronous transcription REST reference describes a stricter serving input: smaller than 250 MB and shorter than two hours. These are separately stated model-card and interface limits. Prepare files within the allowance of the intended interface and organize longer material at meaningful recording boundaries. A file-size limit concerns the encoded audio, while the duration limit concerns the speech recording’s length; neither defines a text-model context window.\n\nThe model card’s sample output is JSON containing the recording duration and combined recognized text. The text can become an accessible transcript, a script for captions or an input for later summarization and search. Caption cue grouping and alignment to the recording are separate preparation when the result does not include the required timing data. For interview records, attributing a line to a named participant needs a separate checked record of who spoke. Preserve the recognized wording and the audio as the basis for those later artifacts.\n\nThe edition’s boundaries matter when planning the output. The 1.5 card explicitly places speaker diarization outside its supported capabilities, and the maker’s version comparison assigns word-level timestamps and clean/verbatim style controls to MAI-Transcribe 2. A native streaming model is also a separate edition. This 1.5 profile therefore centres on recognized text from recordings and its documented domain context, without turning another edition’s speaker, timing or live-input features into part of the result. Review names, numbers, technical terms and unclear speech before preparing a quoted passage, a caption track or a derived document.", "editorialOverlay": { "copy_fields": null, "copy_sha256": null, "copy_source": null, "html_sha256": "01ef601c2a244f107d930f268ce2148ac56d213f5f82ca431dbe6d6c4b49a398", "module_sha256": "ee9dc1c50b1e6a8fb7fa062030fcb4646b9aaaeb08be513b723a5a92a7290f06", "raw_model_unchanged": true, "schema": "ailoot-native-model-editorial-v1" }, "locale": "en", "methods": [ { "html": "microsoft/mai-transcribe-1.5Microsoft AI MAI-Transcribe 1.5 converts recorded speech into text across 43 languages, with automatic language identification and an emphasis on accents and noisy audio. Its model card describes domain-aware keyword biasing for names and specialist terms. Use the transcript as a basis for checked notes or captions.
This description covers the model’s capabilities. Available in Ailoot inputs, outputs and options are listed for the selected method below.
Microsoft AI MAI-Transcribe 1.5 is a speech-to-text model for converting recorded audio into written content. Its exact model card describes recognition across 43 languages, automatic language identification and robustness across accents, dialects, speaking styles and noisy conditions. The edition expands the 25-language set of MAI-Transcribe 1 by eighteen languages. A useful result is the written speech from a lecture, an interview, a call or a voice note, ready to check against the recording before a later editing or analysis step.
Prepare a complete, intelligible recording in one of the maker’s listed formats: WAV, MP3 or FLAC. Keep the original file and any division into sections so the transcript can be compared with the audio. For a lecture, retain the phrases around a technical term rather than extracting only the difficult word. For a voice note, preserve a correction that changes a name, time or number. The surrounding speech provides context for recognizing what was said, and the listener’s review can identify where the output needs correction.
Automatic language identification lets a recording be processed without requiring a language choice in advance. The 43-language set includes English, French, German, Italian, Spanish, Hindi, Chinese, Japanese, Korean, Arabic, Indonesian, Russian, Thai and Vietnamese. The added languages include Bulgarian, Catalan, Greek, Ukrainian and several South Asian languages. Check that the intended language belongs to this edition’s set rather than borrowing the broader coverage of MAI-Transcribe 2. For multilingual collections, retain the recording’s language with the reviewed text so downstream editing follows the actual spoken content.
The 1.5 model card describes keyword or entity biasing with up to 200 terms, and its release explains the purpose: make likely names, acronyms and specialist words available as context. The maker describes contextual hints rather than blindly forcing supplied words into the result. Prepare a relevant vocabulary, such as the project name, a technical instrument or the correct spelling of a person discussed in the recording. Use that terminology as context for the actual recording, and check the recognized terms against the speech.
For a recorded engineering lesson, prepare the names of the instruments discussed and the exact spelling of any project acronym. Write their exact spellings in a focused terminology list and keep the list alongside the audio and transcript. Review those occurrences against the spoken passages rather than treating a vocabulary entry as proof of a match. Names that sound similar can still need checking. The result should preserve the words actually spoken, including a correction or qualification that changes their meaning.
The manufacturer emphasizes performance on real-world audio, including background noise and diverse microphone conditions. Practical review can focus on passages with simultaneous speech, faint voices or words partly covered by another sound. Keep uncertainties tied to the corresponding audio passage instead of filling them from assumptions about the topic. A plausible sentence may still contain the wrong proper noun or number. For a call transcript, the useful deliverable is a checked record of the spoken content that a later analysis can use, rather than a guessed account of the caller’s intent.
The exact 1.5 model card lists a maximum file size of 300 MB and two hours. Its linked synchronous transcription REST reference describes a stricter serving input: smaller than 250 MB and shorter than two hours. These are separately stated model-card and interface limits. Prepare files within the allowance of the intended interface and organize longer material at meaningful recording boundaries. A file-size limit concerns the encoded audio, while the duration limit concerns the speech recording’s length; neither defines a text-model context window.
The model card’s sample output is JSON containing the recording duration and combined recognized text. The text can become an accessible transcript, a script for captions or an input for later summarization and search. Caption cue grouping and alignment to the recording are separate preparation when the result does not include the required timing data. For interview records, attributing a line to a named participant needs a separate checked record of who spoke. Preserve the recognized wording and the audio as the basis for those later artifacts.
The edition’s boundaries matter when planning the output. The 1.5 card explicitly places speaker diarization outside its supported capabilities, and the maker’s version comparison assigns word-level timestamps and clean/verbatim style controls to MAI-Transcribe 2. A native streaming model is also a separate edition. This 1.5 profile therefore centres on recognized text from recordings and its documented domain context, without turning another edition’s speaker, timing or live-input features into part of the result. Review names, numbers, technical terms and unclear speech before preparing a quoted passage, a caption track or a derived document.
POST /v1/audio/transcriptionsaudiotext| Field | Type | Requirement | Values & limits | Default |
|---|---|---|---|---|
Fieldmodel | Typestring | RequirementRequired | Values & limits microsoft/mai-transcribe-1.5 | Default— |
Fieldinput_audio | Typeobject | RequirementRequired | Values & limits | Default— |
Fieldlanguage | Typestring | RequirementOptional | Values & limits | Default— |
Fieldtemperature | Typenumber | RequirementOptional | Values & limits Minimum: 0 Maximum: 1 | Default— |
Fieldresponse_format | Typestring | RequirementOptional | Values & limits json, verbose_json | Defaultjson |
Fieldtimestamp_granularities | Typearray | RequirementOptional | Values & limits | Default— |
Fieldprovider | Typeobject | RequirementOptional | Values & limits | Default— |
| Parameter | Contract |
|---|---|
Parametercontent_type | Contractapplication/json |
Parameterrequired | Contractmodel, input_audio |
Parametermaterial_types | Contractaudio |
Parametermaterial_conditions | ContractRead an actual WAV file and encode its real bytes for the declared JSON input_audio; no multipart route is advertised for this exact JSON-only declaration. |
Lifecycle: synchronous_result_then_owned_library
GET /v1/media/tasks/{task_id} · actual returned task_id or media library record
GET /v1/media/assets/{asset_id}/content
GET /v1/media/assets
GET /v1/media/assets/{asset_id}
The displayed rates estimate the cost before generation. The final charge is calculated from the cost recorded for the original generation. Measured output duration and forecast quantities are not the final billable quantity.
Tariff references do not add request options. Supported parameters are listed for the selected method.
Match each charge to the conditions shown. Use one rate for each set of alternative configurations.
| Configuration | Rate schedule | Forecast quantity | Billing unit | Charge |
|---|---|---|---|---|
| ConfigurationBase token rates | Rate scheduleBase rate | Forecast quantity— | Billing unit— | Charge Input · Input tokens · Estimate for comparison only — / 1K tokens Output · Output tokens · Estimate for comparison only — / 1K tokens Input · Cache read · Estimate for comparison only — / 1K tokens Input · Cache write · Estimate for comparison only — / 1K tokens Input · Cache write, 1 hour · Estimate for comparison only — / 1K tokens |
| Configurationinput audio hour | Rate schedule— | Forecast quantity— | Billing unitper input audio hour | Charge Input · input audio hour · Estimate for comparison only $0.72 per input audio hour |
| ConfigurationEstimated generation cost | Rate schedule— | Forecast quantityone generation estimate | Billing unit— | Charge Output · Estimated generation cost · Estimate for comparison only $0.0002 estimate before generation This estimate is not an additional charge or a fixed rate for the completed generation. |
Ailoot reserves an estimated amount before a request. After the result, it settles the final charge from measured usage. Streaming settlement uses the final usage record. How reservation and settlement work →
set -eu\n: "${AILOOT_API_KEY:?Set your Ailoot API key}"\n: "${AILOOT_AUDIO_FILE:?Set the path to your actual WAV recording}"\npython3 - <<'PY_REQUEST'\nimport json,os\nfrom pathlib import Path\nimport base64\naudio=Path(os.environ['AILOOT_AUDIO_FILE']).read_bytes()\nif not (audio[:4]==b'RIFF' and audio[8:12]==b'WAVE'): raise SystemExit('Supply an actual WAV audio file')\nif len(audio)>26214400: raise SystemExit('Use a documented supported large-file flow')\nrequest={'model': 'microsoft/mai-transcribe-1.5', 'input_audio': {'data': '${BASE64_ACTUAL_WAV_BYTES}', 'format': 'wav'}}\nrequest['input_audio']['data']=base64.b64encode(audio).decode('ascii')\nPath('request.json').write_text(json.dumps(request,ensure_ascii=False),encoding='utf-8')\nPY_REQUEST\ncurl --fail-with-body --silent --show-error https://api.ailoot.ai/v1/audio/transcriptions \\\n -H "Authorization: Bearer $AILOOT_API_KEY" -H 'Content-Type: application/json' \\\n --data-binary @request.json -D response-headers.txt -o response.bin\n# Keep the complete original response and headers. Read every returned segment, word and usage field.\n# Preserve actual operation/task/asset IDs; continue reads of the original result without resubmitting.\nGET /v1/media/tasks/{task_id}GET /v1/media/assets/{asset_id}/contentGET /v1/media/assetsGET /v1/media/assets/{asset_id}set -eu\n: "${AILOOT_API_KEY:?Set your Ailoot API key}"\ncurl --fail-with-body --silent --show-error https://api.ailoot.ai/v1/media/assets \\\n -H "Authorization: Bearer $AILOOT_API_KEY" -o owned-assets.json\n# Preserve every returned asset and follow its exact signed content URL. Do not construct a signed token.\nKeep every complete returned response, file and content part.
Retain the complete response and headers, every actual output, original returned identifiers and all usage fields. Recover the original operation without submitting creation again.
Keep the returned X-Client-Request-ID for reconciliation. Read only actual returned task_id/asset_id values or the owned media library; never build a task URL from the operation ID and never resubmit creation.
| Field | Type |
|---|---|
Fieldtext | Type |
Fieldusage | Type |
Fieldsegments[] | Type |
Fieldwords[] | Type |
Fieldspeaker fields | Type |
Fieldtimestamps | Type |
Ailoot Ultra microsoft/mai-transcribe-1.5
Use the method paths shown above.