Voxtral TTS is Mistral AI's 4B text-to-speech model for generating spoken audio from text and a voice reference. The maker's model page identifies the 2603 edition as voxtral-mini-tts-2603, while the released weight card is named Voxtral 4B TTS 2603. This entry concerns that speech-generation release, distinct from Voxtral Small audio understanding, transcription and later real-time input models.
Expressive speech and voice adaptation
The model is designed to capture a reference voice's pauses, rhythm, intonation, accent and emotional rendering. The voice acts as an instruction for how the text should sound; the maker's overview does not require separate prosody or emotion tags, or a transcript of the reference recording. The source text determines what is said, while the audio reference guides its delivery.
The overview describes adapting a voice from as little as two to three seconds of audio. The architecture description discusses voice prompts of five to twenty-five seconds. Those statements have different scopes: a short adaptation example is not a requirement to use the shortest possible reference. Choose clear speech that represents the style and accent wanted for the final result.
The released card includes twenty preset voices and adaptation to new voices. The maker's speech workflow can use a saved voice profile or a one-off reference clip. Saving a profile allows the same voice reference to be reused across separate speech requests, while a one-off clip supplies it for the current task. A profile and a generated recording are different artifacts.
Languages and cross-language delivery
The release supports English, French, Spanish, German, Italian, Portuguese, Dutch, Arabic and Hindi, with dialect variation, cross-lingual voice transfer and code-mixing described by the maker. For the best result, the speech guide recommends matching the reference language to the text language. Cross-language transfer can preserve the reference's accent while speaking the prepared text in another supported language.
A practical narration task can use the same voice style for an explanation in different supported languages, with each text prepared for the intended listener. A voice-assistant reply can use a short, conversational script rather than reading an unedited document aloud. The text-to-speech model renders the prepared words; any translation or answer-writing stage that produces those words is a separate step.
Preparing the text
The maker's speech guide recommends verbalizable text. Write numbers and symbols in the form that should be spoken so the model does not need to choose between possible readings. A number can be a quantity, time or identifier; define the intended spoken wording before synthesis. Spell out abbreviations when pronunciation matters, using separated letters where that is the intended reading.
Avoid rich formatting, emojis and special characters that will not be rendered as spoken content. Prepare complete sentences with the punctuation and wording needed for the delivery. The guide recommends prompts below 300 words for best results. This is a quality recommendation rather than a declared hard maximum input length.
For a procedural explanation, write the action and the condition in a sentence a listener can follow. For an announcement, place the important information before optional detail. For a voice reply, retain the factual answer and remove display-only formatting or source notation that should not be spoken. These preparations make the script itself suitable for audio before the reference determines its style.
Streaming, output and duration
The maker documents basic and streaming delivery and describes native generation of up to two minutes of audio. The released weight card also lists batch inference for preparing multiple speech tasks. Its longer service generations use interleaving rather than treating one native chunk as unbounded. For longer narration, prepare coherent sections and keep their voice and text preparation consistent with the delivery workflow that handles continuation.
The released card specifies 24 kHz audio and names WAV, PCM, FLAC, MP3, AAC and Opus output formats. The speech-service guide discusses MP3 for compressed delivery, WAV for uncompressed PCM, FLAC for lossless compression and PCM or Opus for streaming. Choose a format supported by the connected workflow and suitable for its player or processing step. Format selection and streaming behavior are distinct from the voice adaptation itself.
Model architecture and practical boundaries
The announcement describes an autoregressive, flow-matching architecture with a 3.4B Transformer decoder backbone, a 390-million-parameter acoustic Transformer and a 300-million-parameter neural audio codec. The decoder predicts semantic content and the acoustic component produces speech representation, while the codec handles audio. These components explain the approximate 4B model label; they are not extra input languages or a different answer-generation model.
A practical output is the synthesized recording or audio stream associated with its prepared script and chosen voice. Keep those associations when generating several sections so the application can save, play or assemble the correct material. The model produces audio rather than a transcript, semantic-search vector or newly authored factual answer.
Preserve the 2603 speech edition, supported language and reference style. The read sources provide the native two-minute statement and the under-300-word preparation recommendation, but do not establish a separate hard maximum text-character limit or total stitched-recording duration. Use the actual text, voice and delivery requirements to define the finished audio artifact.
", "aboutText": "Voxtral TTS is Mistral AI's 4B text-to-speech model for generating spoken audio from text and a voice reference. The maker's model page identifies the 2603 edition as voxtral-mini-tts-2603, while the released weight card is named Voxtral 4B TTS 2603. This entry concerns that speech-generation release, distinct from Voxtral Small audio understanding, transcription and later real-time input models.\n\nExpressive speech and voice adaptation\n\nThe model is designed to capture a reference voice's pauses, rhythm, intonation, accent and emotional rendering. The voice acts as an instruction for how the text should sound; the maker's overview does not require separate prosody or emotion tags, or a transcript of the reference recording. The source text determines what is said, while the audio reference guides its delivery.\n\nThe overview describes adapting a voice from as little as two to three seconds of audio. The architecture description discusses voice prompts of five to twenty-five seconds. Those statements have different scopes: a short adaptation example is not a requirement to use the shortest possible reference. Choose clear speech that represents the style and accent wanted for the final result.\n\nThe released card includes twenty preset voices and adaptation to new voices. The maker's speech workflow can use a saved voice profile or a one-off reference clip. Saving a profile allows the same voice reference to be reused across separate speech requests, while a one-off clip supplies it for the current task. A profile and a generated recording are different artifacts.\n\nLanguages and cross-language delivery\n\nThe release supports English, French, Spanish, German, Italian, Portuguese, Dutch, Arabic and Hindi, with dialect variation, cross-lingual voice transfer and code-mixing described by the maker. For the best result, the speech guide recommends matching the reference language to the text language. Cross-language transfer can preserve the reference's accent while speaking the prepared text in another supported language.\n\nA practical narration task can use the same voice style for an explanation in different supported languages, with each text prepared for the intended listener. A voice-assistant reply can use a short, conversational script rather than reading an unedited document aloud. The text-to-speech model renders the prepared words; any translation or answer-writing stage that produces those words is a separate step.\n\nPreparing the text\n\nThe maker's speech guide recommends verbalizable text. Write numbers and symbols in the form that should be spoken so the model does not need to choose between possible readings. A number can be a quantity, time or identifier; define the intended spoken wording before synthesis. Spell out abbreviations when pronunciation matters, using separated letters where that is the intended reading.\n\nAvoid rich formatting, emojis and special characters that will not be rendered as spoken content. Prepare complete sentences with the punctuation and wording needed for the delivery. The guide recommends prompts below 300 words for best results. This is a quality recommendation rather than a declared hard maximum input length.\n\nFor a procedural explanation, write the action and the condition in a sentence a listener can follow. For an announcement, place the important information before optional detail. For a voice reply, retain the factual answer and remove display-only formatting or source notation that should not be spoken. These preparations make the script itself suitable for audio before the reference determines its style.\n\nStreaming, output and duration\n\nThe maker documents basic and streaming delivery and describes native generation of up to two minutes of audio. The released weight card also lists batch inference for preparing multiple speech tasks. Its longer service generations use interleaving rather than treating one native chunk as unbounded. For longer narration, prepare coherent sections and keep their voice and text preparation consistent with the delivery workflow that handles continuation.\n\nThe released card specifies 24 kHz audio and names WAV, PCM, FLAC, MP3, AAC and Opus output formats. The speech-service guide discusses MP3 for compressed delivery, WAV for uncompressed PCM, FLAC for lossless compression and PCM or Opus for streaming. Choose a format supported by the connected workflow and suitable for its player or processing step. Format selection and streaming behavior are distinct from the voice adaptation itself.\n\nModel architecture and practical boundaries\n\nThe announcement describes an autoregressive, flow-matching architecture with a 3.4B Transformer decoder backbone, a 390-million-parameter acoustic Transformer and a 300-million-parameter neural audio codec. The decoder predicts semantic content and the acoustic component produces speech representation, while the codec handles audio. These components explain the approximate 4B model label; they are not extra input languages or a different answer-generation model.\n\nA practical output is the synthesized recording or audio stream associated with its prepared script and chosen voice. Keep those associations when generating several sections so the application can save, play or assemble the correct material. The model produces audio rather than a transcript, semantic-search vector or newly authored factual answer.\n\nPreserve the 2603 speech edition, supported language and reference style. The read sources provide the native two-minute statement and the under-300-word preparation recommendation, but do not establish a separate hard maximum text-character limit or total stitched-recording duration. Use the actual text, voice and delivery requirements to define the finished audio artifact.", "editorialOverlay": { "copy_fields": null, "copy_sha256": null, "copy_source": null, "html_sha256": "4e636b7acc9e1f34f6b4c6036b70b64315e28689cb65486028251c12dc2a3565", "module_sha256": "ee9dc1c50b1e6a8fb7fa062030fcb4646b9aaaeb08be513b723a5a92a7290f06", "raw_model_unchanged": true, "schema": "ailoot-native-model-editorial-v1" }, "locale": "en", "methods": [ { "html": "mistralai/voxtral-mini-tts-2603Voxtral TTS is Mistral AI's 4B speech-generation model with reference-based voice adaptation across nine languages. It turns prepared text into expressive audio, supports reusable or one-off voice prompts and streaming, and produces 24 kHz output in documented audio formats.
This description covers the model’s capabilities. Available in Ailoot inputs, outputs and options are listed for the selected method below.
Voxtral TTS is Mistral AI's 4B text-to-speech model for generating spoken audio from text and a voice reference. The maker's model page identifies the 2603 edition as voxtral-mini-tts-2603, while the released weight card is named Voxtral 4B TTS 2603. This entry concerns that speech-generation release, distinct from Voxtral Small audio understanding, transcription and later real-time input models.
Expressive speech and voice adaptation
The model is designed to capture a reference voice's pauses, rhythm, intonation, accent and emotional rendering. The voice acts as an instruction for how the text should sound; the maker's overview does not require separate prosody or emotion tags, or a transcript of the reference recording. The source text determines what is said, while the audio reference guides its delivery.
The overview describes adapting a voice from as little as two to three seconds of audio. The architecture description discusses voice prompts of five to twenty-five seconds. Those statements have different scopes: a short adaptation example is not a requirement to use the shortest possible reference. Choose clear speech that represents the style and accent wanted for the final result.
The released card includes twenty preset voices and adaptation to new voices. The maker's speech workflow can use a saved voice profile or a one-off reference clip. Saving a profile allows the same voice reference to be reused across separate speech requests, while a one-off clip supplies it for the current task. A profile and a generated recording are different artifacts.
Languages and cross-language delivery
The release supports English, French, Spanish, German, Italian, Portuguese, Dutch, Arabic and Hindi, with dialect variation, cross-lingual voice transfer and code-mixing described by the maker. For the best result, the speech guide recommends matching the reference language to the text language. Cross-language transfer can preserve the reference's accent while speaking the prepared text in another supported language.
A practical narration task can use the same voice style for an explanation in different supported languages, with each text prepared for the intended listener. A voice-assistant reply can use a short, conversational script rather than reading an unedited document aloud. The text-to-speech model renders the prepared words; any translation or answer-writing stage that produces those words is a separate step.
Preparing the text
The maker's speech guide recommends verbalizable text. Write numbers and symbols in the form that should be spoken so the model does not need to choose between possible readings. A number can be a quantity, time or identifier; define the intended spoken wording before synthesis. Spell out abbreviations when pronunciation matters, using separated letters where that is the intended reading.
Avoid rich formatting, emojis and special characters that will not be rendered as spoken content. Prepare complete sentences with the punctuation and wording needed for the delivery. The guide recommends prompts below 300 words for best results. This is a quality recommendation rather than a declared hard maximum input length.
For a procedural explanation, write the action and the condition in a sentence a listener can follow. For an announcement, place the important information before optional detail. For a voice reply, retain the factual answer and remove display-only formatting or source notation that should not be spoken. These preparations make the script itself suitable for audio before the reference determines its style.
Streaming, output and duration
The maker documents basic and streaming delivery and describes native generation of up to two minutes of audio. The released weight card also lists batch inference for preparing multiple speech tasks. Its longer service generations use interleaving rather than treating one native chunk as unbounded. For longer narration, prepare coherent sections and keep their voice and text preparation consistent with the delivery workflow that handles continuation.
The released card specifies 24 kHz audio and names WAV, PCM, FLAC, MP3, AAC and Opus output formats. The speech-service guide discusses MP3 for compressed delivery, WAV for uncompressed PCM, FLAC for lossless compression and PCM or Opus for streaming. Choose a format supported by the connected workflow and suitable for its player or processing step. Format selection and streaming behavior are distinct from the voice adaptation itself.
Model architecture and practical boundaries
The announcement describes an autoregressive, flow-matching architecture with a 3.4B Transformer decoder backbone, a 390-million-parameter acoustic Transformer and a 300-million-parameter neural audio codec. The decoder predicts semantic content and the acoustic component produces speech representation, while the codec handles audio. These components explain the approximate 4B model label; they are not extra input languages or a different answer-generation model.
A practical output is the synthesized recording or audio stream associated with its prepared script and chosen voice. Keep those associations when generating several sections so the application can save, play or assemble the correct material. The model produces audio rather than a transcript, semantic-search vector or newly authored factual answer.
Preserve the 2603 speech edition, supported language and reference style. The read sources provide the native two-minute statement and the under-300-word preparation recommendation, but do not establish a separate hard maximum text-character limit or total stitched-recording duration. Use the actual text, voice and delivery requirements to define the finished audio artifact.
POST /v1/audio/speechtextaudio| Field | Type | Requirement | Values & limits | Default |
|---|---|---|---|---|
Fieldmodel | Typestring | RequirementRequired | Values & limits mistralai/voxtral-mini-tts-2603 | Default— |
Fieldinput | Typestring | RequirementRequired | Values & limits Minimum length: 1 Unit: Unicode characters | Default— |
Fieldvoice | Typestring | RequirementRequired | Values & limits en_paul_sad, en_paul_neutral, en_paul_happy, en_paul_frustrated, en_paul_excited, en_paul_confident, en_paul_cheerful, en_paul_angry, gb_oliver_neutral, gb_oliver_sad, gb_oliver_excited, gb_oliver_curious, gb_oliver_confident, gb_oliver_cheerful, gb_oliver_angry, gb_jane_sarcasm, gb_jane_confused, gb_jane_shameful, gb_jane_sad, gb_jane_neutral, gb_jane_jealousy, gb_jane_frustrated, gb_jane_curious, gb_jane_confident, fr_marie_sad, fr_marie_neutral, fr_marie_happy, fr_marie_excited, fr_marie_curious, fr_marie_angry | Default— |
Fieldresponse_format | Typestring | RequirementRequired | Values & limits mp3 | Default— |
text_to_speech · Required: input
| Parameter | Contract |
|---|---|
Parametercontent_type | Contractapplication/json |
Parameterrequired | Contractmodel, input, voice, response_format |
Parametervoice_ids | Contracten_paul_sad, en_paul_neutral, en_paul_happy, en_paul_frustrated, en_paul_excited, en_paul_confident, en_paul_cheerful, en_paul_angry, gb_oliver_neutral, gb_oliver_sad, gb_oliver_excited, gb_oliver_curious, gb_oliver_confident, gb_oliver_cheerful, gb_oliver_angry, gb_jane_sarcasm, gb_jane_confused, gb_jane_shameful, gb_jane_sad, gb_jane_neutral, gb_jane_jealousy, gb_jane_frustrated, gb_jane_curious, gb_jane_confident, fr_marie_sad, fr_marie_neutral, fr_marie_happy, fr_marie_excited, fr_marie_curious, fr_marie_angry |
Parameterresponse_format | Contractmp3 |
Parametermax_input_characters | Contractnull |
Parameterother_options | Contractfalse |
Parameterinput | Contractexact UTF-8 text; Unicode character count without normalization |
Parametermaterial_types | Contracttext |
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}
Select exact voice from this model profile.
Input characters and audio seconds are different meters.
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 | Billable quantity | Billing unit | Charge |
|---|---|---|---|---|
| ConfigurationBase token rates | Rate scheduleBase rate | Billable quantity— | 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 |
| ConfigurationUltra | Rate schedule— | Billable quantityEach Unicode code point counts as one character, including Han characters. | Billing unitper 1000 input character | Charge Input · input character $0.032 per 1000 input character |
Model-level reference schedules do not change method selection, quotes or charges. Active charges use the selected Ailoot method rates.
| Configuration | Billing conditions | Charge |
|---|---|---|
| ConfigurationMistral: Voxtral Mini TTS · Default configuration | Billing conditionsDefault configuration | Charge Input · Input characters $0.000032 per input character |
Ailoot reserves an estimated amount before a request. After the result, it settles the final charge from measured usage. Streaming settlement uses the final usage record. How reservation and settlement work →
set -eu\n: "${AILOOT_API_KEY:?Set your Ailoot API key}"\n: "${AILOOT_VOICE_ID:?Choose an accepted voice ID from this exact model profile}"\nexport AILOOT_VOICE_ID\npython3 - <<'PY_REQUEST'\nimport json,os\nfrom pathlib import Path\np=json.loads('{"model": "mistralai/voxtral-mini-tts-2603", "input": "Good afternoon! This is a short voice take."}')\nvoice=os.environ["AILOOT_VOICE_ID"]\nallowed=json.loads('["en_paul_sad", "en_paul_neutral", "en_paul_happy", "en_paul_frustrated", "en_paul_excited", "en_paul_confident", "en_paul_cheerful", "en_paul_angry", "gb_oliver_neutral", "gb_oliver_sad", "gb_oliver_excited", "gb_oliver_curious", "gb_oliver_confident", "gb_oliver_cheerful", "gb_oliver_angry", "gb_jane_sarcasm", "gb_jane_confused", "gb_jane_shameful", "gb_jane_sad", "gb_jane_neutral", "gb_jane_jealousy", "gb_jane_frustrated", "gb_jane_curious", "gb_jane_confident", "fr_marie_sad", "fr_marie_neutral", "fr_marie_happy", "fr_marie_excited", "fr_marie_curious", "fr_marie_angry"]')\nif voice not in allowed: raise SystemExit("Choose an accepted voice ID from this exact model profile")\np["voice"]=voice\np["response_format"]="mp3"\nPath("request.json").write_text(json.dumps(p,ensure_ascii=False))\nPY_REQUEST\ncurl --fail-with-body --silent --show-error https://api.ailoot.ai/v1/audio/speech \\\n -H "Authorization: Bearer $AILOOT_API_KEY" -H 'Content-Type: application/json' \\\n --data-binary @request.json -D response-headers.txt -o speech.mp3\nmp3
GET /v1/media/tasks/{task_id}GET /v1/media/assets/{asset_id}/contentGET /v1/media/assetsGET /v1/media/assets/{asset_id}set -eu\n: "${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 $AILOOT_API_KEY" -o media-assets.json\nset -eu\n: "${AILOOT_API_KEY:?Set your Ailoot API key}"\n: "${AILOOT_MEDIA_TASK_ID:?Set the actual ID returned by the response or media-library metadata}"\nexport AILOOT_MEDIA_TASK_ID\nresource_id=$(python3 -c 'import os,urllib.parse;print(urllib.parse.quote(os.environ["AILOOT_MEDIA_TASK_ID"],safe=""))')\ncurl --fail-with-body --silent --show-error "https://api.ailoot.ai/v1/media/tasks/$resource_id" \\\n -H "Authorization: Bearer $AILOOT_API_KEY" -o media-task.json\npython3 - <<'PY_RESULT'\nimport json\nfrom pathlib import Path\nprint(json.dumps(json.loads(Path("media-task.json").read_text())))\nPY_RESULT\nset -eu\n: "${AILOOT_API_KEY:?Set your Ailoot API key}"\n: "${AILOOT_ASSET_ID:?Set the actual ID returned by the response or media-library metadata}"\nexport AILOOT_ASSET_ID\nresource_id=$(python3 -c 'import os,urllib.parse;print(urllib.parse.quote(os.environ["AILOOT_ASSET_ID"],safe=""))')\ncurl --fail-with-body --silent --show-error "https://api.ailoot.ai/v1/media/assets/$resource_id" \\\n -H "Authorization: Bearer $AILOOT_API_KEY" -o media-asset.json\npython3 - <<'PY_RESULT'\nimport json\nfrom pathlib import Path\nprint(json.dumps(json.loads(Path("media-asset.json").read_text())))\nPY_RESULT\nKeep every complete returned response, file and content part.
Keep the complete binary result and headers. If JSON/multiple outputs are returned, retain every data item and fetch each actual signed asset URL.
Retain original operation_id/X-Client-Request-ID for reconciliation without resubmission; no dedicated audio poll route.
| Field | Type |
|---|---|
Fieldbinary audio | Typebytes |
Fieldresponse headers | Typeheaders |
| Response | Condition | Next step |
|---|---|---|
Response400 · invalid_request_error | ConditionInvalid required fields, material representation or parameter value | Next stepCorrect the indicated fields before a new creation request. |
Responseserver_error | ConditionThe returned response says the operation or result requires a check | Next stepKeep the original returned operation/task ID, request headers and response; continue supported reads of that original operation. Do not resubmit the creation request. |
Ailoot Ultra mistralai/voxtral-mini-tts-2603
Use the method paths shown above.