Voxtral Mini Transcribe 2 is Mistral AI’s transcription-focused Voxtral edition identified in the producer model card as voxtral-mini-2602. It is designed for recording-based batch transcription rather than the broader question-answering role of the original Voxtral Mini 3B model. The input is audio and the result is transcribed text, optionally enriched with speaker labels and timing information. Practical uses include meeting transcripts, interview analysis, multi-party call records and text aligned with recorded media.
The manufacturer documents speaker diarisation and start/end timing so a transcript can show who spoke and when. Word-level timestamps support subtitles, audio search and alignment between a text passage and its position in a recording. The producer guide permits segment or word timestamp granularity. For an interview, a useful artifact is a sequence of attributed passages with timing; for subtitle preparation, word or segment timing can guide a later captioning step rather than forcing that timing to be inferred from plain transcript text.
Context biasing accepts up to 100 words or phrases that guide the model toward the intended spelling of names, technical terms and domain-specific vocabulary. Mistral says this feature is optimised for English and that support in other languages is experimental. Prepare a focused vocabulary list from the recording’s subject instead of an unrelated glossary. An original example would supply the project names and specialised terms likely to appear in a technical discussion, then check those spellings in the completed transcript.
This edition supports thirteen languages: English, Chinese, Hindi, Spanish, Arabic, French, Portuguese, Russian, German, Japanese, Korean, Italian and Dutch. The manufacturer also documents noise robustness and recordings up to three hours in a single request. These belong to Mini Transcribe 2’s recording workflow; the original 2507 audio-understanding model’s context and duration statements describe a different release. Automatic language detection is available, and the guide says explicitly supplying a known audio language can improve transcription accuracy.
In the documented producer workflow, audio can be supplied as an uploaded file, a previously uploaded file identifier or a referenced recording. The accompanying upload demonstration accepts MP3, WAV, M4A, FLAC and OGG recordings; choose one of these formats for the demonstrated upload workflow. Keep the recording decodable and include the relevant speech. For lengthy material, retain its identifier and time origin so any later excerpts or edits can still be traced to the source.
The transcription reference returns the text and detected language, with segments when requested, and separately documents streamed transcription events. Diarisation is optional and defaults to false. The guide notes that timestamp_granularities and an explicit language setting are not currently compatible, so choose the desired timing workflow or the manual language hint instead of combining them. Speaker-labelled transcription, vocabulary biasing and timing should be selected according to the intended result rather than assumed from a basic request.
Overlapping speech is a substantive limitation: Mistral says the model typically transcribes one speaker when speakers overlap. A labelled transcript therefore should not be treated as an exhaustive record of every simultaneous utterance. For a meeting record, review difficult overlaps and important names or figures against the recording. The model’s purpose is transcription; a summary, translation or analysis can be a separate task built on that transcript. Voxtral Realtime has a separate live-transcription latency profile, while speech generation belongs to other models.
Prepare the result for its next use. A meeting workflow can preserve the transcript, speaker labels and timing before a separate stage extracts decisions; a media workflow can use the timing to locate passages and draft captions; a terminology-heavy recording can include the focused context-bias list and a review of the resulting spellings. The useful deliverable is text tied to its source recording and configured metadata, with this numbered transcription edition kept distinct from the original Mini model and moving latest names.
", "aboutText": "Voxtral Mini Transcribe 2 is Mistral AI’s transcription-focused Voxtral edition identified in the producer model card as voxtral-mini-2602. It is designed for recording-based batch transcription rather than the broader question-answering role of the original Voxtral Mini 3B model. The input is audio and the result is transcribed text, optionally enriched with speaker labels and timing information. Practical uses include meeting transcripts, interview analysis, multi-party call records and text aligned with recorded media.\n\nThe manufacturer documents speaker diarisation and start/end timing so a transcript can show who spoke and when. Word-level timestamps support subtitles, audio search and alignment between a text passage and its position in a recording. The producer guide permits segment or word timestamp granularity. For an interview, a useful artifact is a sequence of attributed passages with timing; for subtitle preparation, word or segment timing can guide a later captioning step rather than forcing that timing to be inferred from plain transcript text.\n\nContext biasing accepts up to 100 words or phrases that guide the model toward the intended spelling of names, technical terms and domain-specific vocabulary. Mistral says this feature is optimised for English and that support in other languages is experimental. Prepare a focused vocabulary list from the recording’s subject instead of an unrelated glossary. An original example would supply the project names and specialised terms likely to appear in a technical discussion, then check those spellings in the completed transcript.\n\nThis edition supports thirteen languages: English, Chinese, Hindi, Spanish, Arabic, French, Portuguese, Russian, German, Japanese, Korean, Italian and Dutch. The manufacturer also documents noise robustness and recordings up to three hours in a single request. These belong to Mini Transcribe 2’s recording workflow; the original 2507 audio-understanding model’s context and duration statements describe a different release. Automatic language detection is available, and the guide says explicitly supplying a known audio language can improve transcription accuracy.\n\nIn the documented producer workflow, audio can be supplied as an uploaded file, a previously uploaded file identifier or a referenced recording. The accompanying upload demonstration accepts MP3, WAV, M4A, FLAC and OGG recordings; choose one of these formats for the demonstrated upload workflow. Keep the recording decodable and include the relevant speech. For lengthy material, retain its identifier and time origin so any later excerpts or edits can still be traced to the source.\n\nThe transcription reference returns the text and detected language, with segments when requested, and separately documents streamed transcription events. Diarisation is optional and defaults to false. The guide notes that timestamp_granularities and an explicit language setting are not currently compatible, so choose the desired timing workflow or the manual language hint instead of combining them. Speaker-labelled transcription, vocabulary biasing and timing should be selected according to the intended result rather than assumed from a basic request.\n\nOverlapping speech is a substantive limitation: Mistral says the model typically transcribes one speaker when speakers overlap. A labelled transcript therefore should not be treated as an exhaustive record of every simultaneous utterance. For a meeting record, review difficult overlaps and important names or figures against the recording. The model’s purpose is transcription; a summary, translation or analysis can be a separate task built on that transcript. Voxtral Realtime has a separate live-transcription latency profile, while speech generation belongs to other models.\n\nPrepare the result for its next use. A meeting workflow can preserve the transcript, speaker labels and timing before a separate stage extracts decisions; a media workflow can use the timing to locate passages and draft captions; a terminology-heavy recording can include the focused context-bias list and a review of the resulting spellings. The useful deliverable is text tied to its source recording and configured metadata, with this numbered transcription edition kept distinct from the original Mini model and moving latest names.", "editorialOverlay": { "copy_fields": null, "copy_sha256": null, "copy_source": null, "html_sha256": "47acb7bf5b538e1b74d3de4b27a12cce6e2e704178551c71ac30f31d21f2f74a", "module_sha256": "ee9dc1c50b1e6a8fb7fa062030fcb4646b9aaaeb08be513b723a5a92a7290f06", "raw_model_unchanged": true, "schema": "ailoot-native-model-editorial-v1" }, "locale": "en", "methods": [ { "html": "mistralai/voxtral-mini-transcribeVoxtral Mini Transcribe 2 converts recordings into text with speaker diarisation, word-level timestamps and vocabulary biasing. Mistral documents 13 languages and recordings up to three hours for this transcription-focused edition, distinct from Voxtral Realtime.
This description covers the model’s capabilities. Available in Ailoot inputs, outputs and options are listed for the selected method below.
Voxtral Mini Transcribe 2 is Mistral AI’s transcription-focused Voxtral edition identified in the producer model card as voxtral-mini-2602. It is designed for recording-based batch transcription rather than the broader question-answering role of the original Voxtral Mini 3B model. The input is audio and the result is transcribed text, optionally enriched with speaker labels and timing information. Practical uses include meeting transcripts, interview analysis, multi-party call records and text aligned with recorded media.
The manufacturer documents speaker diarisation and start/end timing so a transcript can show who spoke and when. Word-level timestamps support subtitles, audio search and alignment between a text passage and its position in a recording. The producer guide permits segment or word timestamp granularity. For an interview, a useful artifact is a sequence of attributed passages with timing; for subtitle preparation, word or segment timing can guide a later captioning step rather than forcing that timing to be inferred from plain transcript text.
Context biasing accepts up to 100 words or phrases that guide the model toward the intended spelling of names, technical terms and domain-specific vocabulary. Mistral says this feature is optimised for English and that support in other languages is experimental. Prepare a focused vocabulary list from the recording’s subject instead of an unrelated glossary. An original example would supply the project names and specialised terms likely to appear in a technical discussion, then check those spellings in the completed transcript.
This edition supports thirteen languages: English, Chinese, Hindi, Spanish, Arabic, French, Portuguese, Russian, German, Japanese, Korean, Italian and Dutch. The manufacturer also documents noise robustness and recordings up to three hours in a single request. These belong to Mini Transcribe 2’s recording workflow; the original 2507 audio-understanding model’s context and duration statements describe a different release. Automatic language detection is available, and the guide says explicitly supplying a known audio language can improve transcription accuracy.
In the documented producer workflow, audio can be supplied as an uploaded file, a previously uploaded file identifier or a referenced recording. The accompanying upload demonstration accepts MP3, WAV, M4A, FLAC and OGG recordings; choose one of these formats for the demonstrated upload workflow. Keep the recording decodable and include the relevant speech. For lengthy material, retain its identifier and time origin so any later excerpts or edits can still be traced to the source.
The transcription reference returns the text and detected language, with segments when requested, and separately documents streamed transcription events. Diarisation is optional and defaults to false. The guide notes that timestamp_granularities and an explicit language setting are not currently compatible, so choose the desired timing workflow or the manual language hint instead of combining them. Speaker-labelled transcription, vocabulary biasing and timing should be selected according to the intended result rather than assumed from a basic request.
Overlapping speech is a substantive limitation: Mistral says the model typically transcribes one speaker when speakers overlap. A labelled transcript therefore should not be treated as an exhaustive record of every simultaneous utterance. For a meeting record, review difficult overlaps and important names or figures against the recording. The model’s purpose is transcription; a summary, translation or analysis can be a separate task built on that transcript. Voxtral Realtime has a separate live-transcription latency profile, while speech generation belongs to other models.
Prepare the result for its next use. A meeting workflow can preserve the transcript, speaker labels and timing before a separate stage extracts decisions; a media workflow can use the timing to locate passages and draft captions; a terminology-heavy recording can include the focused context-bias list and a review of the resulting spellings. The useful deliverable is text tied to its source recording and configured metadata, with this numbered transcription edition kept distinct from the original Mini model and moving latest names.
POST /v1/audio/transcriptionsaudiotext| Field | Type | Requirement | Values & limits | Default |
|---|---|---|---|---|
Fieldmodel | Typestring | RequirementRequired | Values & limits mistralai/voxtral-mini-transcribe | Default— |
Fieldinput_audio | Typeobject | RequirementRequired | Values & limits | Default— |
| Parameter | Contract |
|---|---|
Parametercontent_type | Contractapplication/json |
Parameterrequired | Contractmodel, input_audio |
Parametermaterial_types | Contractaudio |
Parametermaterial_conditions | ContractThe example reads your actual nonempty PCM WAV file, encodes the complete file bytes as base64 and sends input_audio.format as wav. |
Lifecycle: synchronous_result_then_owned_library
GET /v1/media/tasks/{task_id} · exact task_id from an existing owned media record
GET /v1/media/assets/{asset_id}/content
GET /v1/media/assets
GET /v1/media/assets/{asset_id}
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 | Billing unit | Charge |
|---|---|---|---|
| ConfigurationBase token rates | Rate scheduleBase rate | Billing unit— | Charge Input · Input tokens — / 1K tokens Output · Output tokens — / 1K tokens Input · Cache read — / 1K tokens Input · Cache write — / 1K tokens Input · Cache write, 1 hour — / 1K tokens |
| Configurationinput audio second | Rate schedule— | Billing unitper input audio second | Charge Input · input audio second $0.0001 per input audio second |
Ailoot reserves an estimated amount before a request. After the result, it settles the final charge from measured usage. Streaming settlement uses the final usage record. How reservation and settlement work →
set -eu\n: "${AILOOT_API_KEY:?Set your Ailoot API key}"\n: "${AILOOT_AUDIO_FILE:?Set the path to your actual PCM WAV recording}"\nexport AILOOT_AUDIO_FILE\npython3 - <<'PY_REQUEST'\nimport base64, io, json, os, wave\nfrom pathlib import Path\naudio = Path(os.environ['AILOOT_AUDIO_FILE']).read_bytes()\nif not (audio[:4] == b'RIFF' and audio[8:12] == b'WAVE'):\n raise SystemExit('Supply your actual PCM WAV audio file')\nwith wave.open(io.BytesIO(audio), 'rb') as recording:\n if recording.getcomptype() != 'NONE' or recording.getnframes() <= 0:\n raise SystemExit('Supply a nonempty PCM WAV recording')\nrequest = {'model': 'mistralai/voxtral-mini-transcribe', 'input_audio': {'data': base64.b64encode(audio).decode('ascii'), 'format': 'wav'}}\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# Retain the complete original response, headers and actual X-Client-Request-ID.\n# Read existing owned records for recovery; do not resubmit an uncertain operation.\nComplete original response body, status and forwarded headers. No universal JSON shape, text field or task envelope is added by this handler.
GET /v1/media/tasks/{task_id}GET /v1/media/assets/{asset_id}/contentGET /v1/media/assetsGET /v1/media/assets/{asset_id}Keep 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.
Retain the actual X-Client-Request-ID and complete response. Read the original owned assets and their exact signed content URLs; never resubmit an uncertain creation.
| Field | Type |
|---|---|
Fieldtext (when present) | Type |
Fieldusage (when present) | Type |
| Response | Condition | Next step |
|---|---|---|
Response400 · invalid_request_error | ConditionThe JSON body or explicit canonical model ID is invalid. | Next stepRetain the response and correct the indicated request field. |
Response401 · authentication_error | ConditionThe Ailoot API key is invalid. | Next stepUse your authorized Ailoot API key. |
Response403 · permission_error | ConditionThe key group does not admit this method. | Next stepRetain the returned error details. |
Response404 · not_found_error | ConditionThe selected method has no admitted route. | Next stepRetain the returned error details. |
Response422 · price_unavailable | ConditionThe accepted tariff cannot obtain its required input quantity or bound. | Next stepRetain the response; use the input representation required by the accepted tariff. |
Response503 · | ConditionRequired storage, route, method price or reservation is unavailable. | Next stepKeep the complete original response. An uncertain reservation is not permission to resubmit. |
Response502 · server_error | ConditionThe original result, usage or settlement requires checking. | Next stepRetain the actual request ID and original response. Read the existing owned record; do not resubmit creation. |
Ailoot Ultra mistralai/voxtral-mini-transcribe
Use the method paths shown above.