Speech 2.6 HD is intended for natural speech synthesis and scenarios where beginning a response quickly matters. The manufacturer focuses on intonation, speech fluency and text processing before delivery. The HD variant emphasizes audio quality; this card's input is text and output is speech audio.
The Speech 2.6 generation can process forms often found in working messages: URLs, email addresses, phone numbers, dates and monetary amounts. The model converts them into spoken forms so a line remains understandable by ear. The manufacturer describes these tasks in several languages.
Speech 2.6 provides Fluent LoRA, a fluency enhancement mode. It is intended for synthesis using a prepared voice: it helps retain timbre while producing more connected speech, even if the original sample was uneven or accented. Preparing a cloned voice and using it for narration are different actions.
40 languages are supported, including Russian. Alongside seven basic emotional states, Speech 2.6 provides the fluent-speech option fluent and whisper option whisper. If no emotion is specified, the model chooses it from the text's meaning.
MiniMax provides voice selection and control over pace, volume and pitch. Speed ranges from 0.5–2, volume is above 0 and up to 10, and pitch adjustment ranges from −12 to 12. Emotion can be left to the model according to the text's meaning or specified manually: joy, sadness, anger, fear, disgust, surprise or calm.
Pronunciation can be refined using a replacement dictionary and phonetic notation; these provide the International Phonetic Alphabet, Chinese pinyin and Japanese reading variants. Pauses are inserted between spoken passages in the range 0.01–99.99 seconds. Pause duration is specified to two decimal places; markers are placed between spoken passages and not consecutively. These settings help align names, abbreviations and ambiguous words with the intended narration.
Synthesis accepts text of fewer than 10,000 characters. For long text exceeding 3000 characters, the manufacturer recommends streaming output; receiving completed audio in full is also provided. General synthesis settings include MP3, PCM, FLAC, WAV, Opus and G.711 μ-law variants, in mono or stereo; the available formats for a specific request are determined by Ailoot's controls.
According to MiniMax, the generation targets low latency in voice scenarios. This describes the model; actual response time and availability of extended settings depend on the selected Ailoot access method.
Additional audio settings include sample rates of 8, 16, 22.05, 24, 32 and 44.1 kHz. MP3 provides bitrates of 32, 64, 128 or 256 kbps. Voice coloration can also be changed and one effect applied: spacious echo, hall echo, telephone sound or a robotic voice. Voice-modification effects are supported for MP3, WAV and FLAC when receiving the audio in full, and only for MP3 when streaming. G.711 μ-law formats use an 8 kHz sample rate. These capabilities belong to MiniMax's synthesis settings.
The manufacturer provides speech-text synchronization: sentence timestamps, word timestamps or word timestamps during incremental output. This result helps align speech with displayed text. This additional capability requires support for the corresponding response format.
", "aboutText": "### A natural voice for interaction\n\nSpeech 2.6 HD is intended for natural speech synthesis and scenarios where beginning a response quickly matters. The manufacturer focuses on intonation, speech fluency and text processing before delivery. The HD variant emphasizes audio quality; this card's input is text and output is speech audio.\n\n### Reading unconventional written forms\n\nThe Speech 2.6 generation can process forms often found in working messages: URLs, email addresses, phone numbers, dates and monetary amounts. The model converts them into spoken forms so a line remains understandable by ear. The manufacturer describes these tasks in several languages.\n\n### Fluent speech while preserving timbre\n\nSpeech 2.6 provides Fluent LoRA, a fluency enhancement mode. It is intended for synthesis using a prepared voice: it helps retain timbre while producing more connected speech, even if the original sample was uneven or accented. Preparing a cloned voice and using it for narration are different actions.\n\n**40 languages** are supported, including Russian. Alongside seven basic emotional states, Speech 2.6 provides the fluent-speech option `fluent` and whisper option `whisper`. If no emotion is specified, the model chooses it from the text's meaning.\n\n### Delivery settings\n\nMiniMax provides voice selection and control over pace, volume and pitch. Speed ranges from **0.5–2**, volume is above 0 and up to **10**, and pitch adjustment ranges from **−12 to 12**. Emotion can be left to the model according to the text's meaning or specified manually: joy, sadness, anger, fear, disgust, surprise or calm.\n\nPronunciation can be refined using a replacement dictionary and phonetic notation; these provide the International Phonetic Alphabet, Chinese pinyin and Japanese reading variants. Pauses are inserted between spoken passages in the range **0.01–99.99 seconds**. Pause duration is specified to two decimal places; markers are placed between spoken passages and not consecutively. These settings help align names, abbreviations and ambiguous words with the intended narration.\n\n### Text length and audio\n\nSynthesis accepts text of fewer than **10,000 characters**. For long text exceeding 3000 characters, the manufacturer recommends streaming output; receiving completed audio in full is also provided. General synthesis settings include MP3, PCM, FLAC, WAV, Opus and G.711 μ-law variants, in mono or stereo; the available formats for a specific request are determined by Ailoot's controls.\n\nAccording to MiniMax, the generation targets low latency in voice scenarios. This describes the model; actual response time and availability of extended settings depend on the selected Ailoot access method.\n\nAdditional audio settings include sample rates of **8, 16, 22.05, 24, 32 and 44.1 kHz**. MP3 provides bitrates of **32, 64, 128 or 256 kbps**. Voice coloration can also be changed and one effect applied: spacious echo, hall echo, telephone sound or a robotic voice. Voice-modification effects are supported for MP3, WAV and FLAC when receiving the audio in full, and only for MP3 when streaming. G.711 μ-law formats use an 8 kHz sample rate. These capabilities belong to MiniMax's synthesis settings.\n\nThe manufacturer provides speech-text synchronization: sentence timestamps, word timestamps or word timestamps during incremental output. This result helps align speech with displayed text. This additional capability requires support for the corresponding response format.", "editorialOverlay": { "copy_fields": null, "copy_sha256": null, "copy_source": null, "html_sha256": "bf003a41400110df84ccddd6562ad0675884116a6e331739b32ac8718015ecbf", "module_sha256": "ee9dc1c50b1e6a8fb7fa062030fcb4646b9aaaeb08be513b723a5a92a7290f06", "raw_model_unchanged": true, "schema": "ailoot-native-model-editorial-v1" }, "locale": "en", "methods": [ { "html": "speech-2.6-hdSpeech 2.6 HD is a high-quality speech synthesis model focused on naturalness and voice similarity.
This description covers the model’s capabilities. Available in Ailoot inputs, outputs and options are listed for the selected method below.
Speech 2.6 HD is intended for natural speech synthesis and scenarios where beginning a response quickly matters. The manufacturer focuses on intonation, speech fluency and text processing before delivery. The HD variant emphasizes audio quality; this card's input is text and output is speech audio.
The Speech 2.6 generation can process forms often found in working messages: URLs, email addresses, phone numbers, dates and monetary amounts. The model converts them into spoken forms so a line remains understandable by ear. The manufacturer describes these tasks in several languages.
Speech 2.6 provides Fluent LoRA, a fluency enhancement mode. It is intended for synthesis using a prepared voice: it helps retain timbre while producing more connected speech, even if the original sample was uneven or accented. Preparing a cloned voice and using it for narration are different actions.
40 languages are supported, including Russian. Alongside seven basic emotional states, Speech 2.6 provides the fluent-speech option fluent and whisper option whisper. If no emotion is specified, the model chooses it from the text's meaning.
MiniMax provides voice selection and control over pace, volume and pitch. Speed ranges from 0.5–2, volume is above 0 and up to 10, and pitch adjustment ranges from −12 to 12. Emotion can be left to the model according to the text's meaning or specified manually: joy, sadness, anger, fear, disgust, surprise or calm.
Pronunciation can be refined using a replacement dictionary and phonetic notation; these provide the International Phonetic Alphabet, Chinese pinyin and Japanese reading variants. Pauses are inserted between spoken passages in the range 0.01–99.99 seconds. Pause duration is specified to two decimal places; markers are placed between spoken passages and not consecutively. These settings help align names, abbreviations and ambiguous words with the intended narration.
Synthesis accepts text of fewer than 10,000 characters. For long text exceeding 3000 characters, the manufacturer recommends streaming output; receiving completed audio in full is also provided. General synthesis settings include MP3, PCM, FLAC, WAV, Opus and G.711 μ-law variants, in mono or stereo; the available formats for a specific request are determined by Ailoot's controls.
According to MiniMax, the generation targets low latency in voice scenarios. This describes the model; actual response time and availability of extended settings depend on the selected Ailoot access method.
Additional audio settings include sample rates of 8, 16, 22.05, 24, 32 and 44.1 kHz. MP3 provides bitrates of 32, 64, 128 or 256 kbps. Voice coloration can also be changed and one effect applied: spacious echo, hall echo, telephone sound or a robotic voice. Voice-modification effects are supported for MP3, WAV and FLAC when receiving the audio in full, and only for MP3 when streaming. G.711 μ-law formats use an 8 kHz sample rate. These capabilities belong to MiniMax's synthesis settings.
The manufacturer provides speech-text synchronization: sentence timestamps, word timestamps or word timestamps during incremental output. This result helps align speech with displayed text. This additional capability requires support for the corresponding response format.
POST /v1/audio/speechtextspeech| Field | Type | Requirement | Values & limits | Default |
|---|---|---|---|---|
Fieldmodel | Typestring | RequirementRequired | Values & limits speech-2.6-hd | Default— |
Fieldinput | Typestring | RequirementRequired | Values & limits Minimum length: 1 Maximum length (exclusive): 10,000 Unit: Unicode characters | Default— |
Fieldvoice | Typestring | RequirementRequired | Values & limits Pattern: ^[A-Za-z0-9_-]{1,128}$ | Default— |
Fieldresponse_format | Typestring | RequirementOptional | Values & limits mp3 | Default— |
text_to_speech · Required: input
| Parameter | Contract |
|---|---|
Parametermaterial_types | Contracttext |
Parameterrequired | Contractnull |
Parameteroptional_controls | Contractnull |
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}
Method shapes and IDs are exact; supplier capability is separate from admission.
POST /v1/audio/speechAiloot 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_TEXT_FILE:?Set your actual UTF-8 narration script}"\nexport AILOOT_TEXT_FILE\n: "${AILOOT_VOICE_ID:?Set the actual accepted voice ID for this model}"\nexport AILOOT_VOICE_ID\npython3 - <<'PY_REQUEST'\nimport json,os\nfrom pathlib import Path\np=json.loads('{"model": "speech-2.6-hd", "input": ""}')\nf=Path(os.environ["AILOOT_TEXT_FILE"])\nif not f.is_file():raise SystemExit("Choose your actual UTF-8 source material")\ntext=f.read_text(encoding="utf-8")\nif not text.strip():raise SystemExit("Source material must contain text")\np["input"]=text\nimport re\nvoice=os.environ["AILOOT_VOICE_ID"]\nif not re.fullmatch(r"[A-Za-z0-9_-]{1,128}",voice):raise SystemExit("Set an actual accepted voice ID")\np["voice"]=voice\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}Keep 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 speech-2.6-hd
Use the method paths shown above.