Create speech · DeutschlandGPT

Generates audio from a text prompt. Compatible with the OpenAI /v1/audio/speech API.

Returns a binary audio stream in the requested response_format. Use /v2/models/all to discover available TTS models. Supported voices and response formats are validated per model — an unsupported value returns a 400 listing the accepted options.

POSThttps://api.deutschlandgpt.de/v2/audio/speechTry it

Example request

curl https://api.deutschlandgpt.de/v2/audio/speech \
  -X POST \
  -H "Authorization: Bearer $DGPT_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
  "model": "gpt-4o-mini-tts",
  "input": "Hallo, wie geht es Ihnen heute?",
  "voice": "nova",
  "response_format": "mp3",
  "speed": 1,
  "instructions": "string",
  "language_code": "de",
  "voice_settings": {},
  "parameters": {}
}'

curl https://api.deutschlandgpt.de/v2/audio/speech \
  -X POST \
  -H "Authorization: Bearer $DGPT_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
  "model": "gpt-4o-mini-tts",
  "input": "Hallo, wie geht es Ihnen heute?",
  "voice": "nova",
  "response_format": "mp3",
  "speed": 1,
  "instructions": "string",
  "language_code": "de",
  "voice_settings": {},
  "parameters": {}
}'

import os, requests

response = requests.post(
    "https://api.deutschlandgpt.de/v2/audio/speech",
    headers={"Authorization": f"Bearer {os.environ['DGPT_API_KEY']}"},
    json={
        "model": "gpt-4o-mini-tts",
        "input": "Hallo, wie geht es Ihnen heute?",
        "voice": "nova",
        "response_format": "mp3",
        "speed": 1,
        "instructions": "string",
        "language_code": "de",
        "voice_settings": {},
        "parameters": {}
    },
)

print(response.json())

import os, requests

response = requests.post(
    "https://api.deutschlandgpt.de/v2/audio/speech",
    headers={"Authorization": f"Bearer {os.environ['DGPT_API_KEY']}"},
    json={
        "model": "gpt-4o-mini-tts",
        "input": "Hallo, wie geht es Ihnen heute?",
        "voice": "nova",
        "response_format": "mp3",
        "speed": 1,
        "instructions": "string",
        "language_code": "de",
        "voice_settings": {},
        "parameters": {}
    },
)

print(response.json())

const response = await fetch('https://api.deutschlandgpt.de/v2/audio/speech', {
  method: 'POST',
  headers: {
    Authorization: `Bearer ${process.env.DGPT_API_KEY}`,
    'Content-Type': 'application/json',
  },
  body: JSON.stringify({
    "model": "gpt-4o-mini-tts",
    "input": "Hallo, wie geht es Ihnen heute?",
    "voice": "nova",
    "response_format": "mp3",
    "speed": 1,
    "instructions": "string",
    "language_code": "de",
    "voice_settings": {},
    "parameters": {}
  }),
});

console.log(await response.json());

const response = await fetch('https://api.deutschlandgpt.de/v2/audio/speech', {
  method: 'POST',
  headers: {
    Authorization: `Bearer ${process.env.DGPT_API_KEY}`,
    'Content-Type': 'application/json',
  },
  body: JSON.stringify({
    "model": "gpt-4o-mini-tts",
    "input": "Hallo, wie geht es Ihnen heute?",
    "voice": "nova",
    "response_format": "mp3",
    "speed": 1,
    "instructions": "string",
    "language_code": "de",
    "voice_settings": {},
    "parameters": {}
  }),
});

console.log(await response.json());

Request body

modelstringrequired

Model to use. Use /v2/models to list TTS models.

inputstringrequired

Text to synthesize. Per-model character limit (4096 for gpt-4o-mini-tts).

voicestringrequired

Voice to use. Optional — defaults to the model's default voice. For gpt-4o-mini-tts: alloy, ash, ballad, coral, echo, fable, nova, onyx, sage, shimmer, verse, marin, cedar. For ElevenLabs models (eleven_multilingual_v2, eleven_flash_v2_5, eleven_v3) it is any voice_id from your ElevenLabs voice library.

response_formatstringoptional

Audio format. Defaults to the model's default format (mp3). ElevenLabs supports mp3, opus, wav, pcm (not aac/flac).

speednumberoptional

Speech rate multiplier.

instructionsstringoptional

Free-text steering for tone and style. Only applied for models that support it (e.g. gpt-4o-mini-tts).

language_codestringoptional

ISO 639-1 language code to enforce (ElevenLabs only).

voice_settingsobjectoptional

Fine-grained voice controls (ElevenLabs only). Ignored by OpenAI/Azure models.

stabilitynumberoptional

similarity_boostnumberoptional

stylenumberoptional

use_speaker_boostbooleanoptional

parametersobjectoptional

Provider-specific passthrough (ElevenLabs only). Unknown keys are ignored; OpenAI/Azure models ignore this field entirely.

seedintegeroptional

Deterministic sampling seed.

previous_textstringoptional

Text that preceded this request, for prosody continuity across stitched generations.

next_textstringoptional

Text that follows this request, for prosody continuity.

previous_request_idsstring[]optional

Up to 3 request_ids of prior generations, for continuity when stitching.

next_request_idsstring[]optional

Up to 3 request_ids of following generations.

apply_text_normalizationstringoptional

Control number/date spelling-out. Default auto.

apply_language_text_normalizationbooleanoptional

Language-specific normalization (e.g. Japanese). Increases latency.

use_pvc_as_ivcbooleanoptional

Use the IVC version of a voice instead of its PVC version.

optimize_streaming_latencyintegeroptional

Latency-vs-quality tradeoff (0 = best quality, 4 = lowest latency).

Response

200audio/mpeg, audio/opus, audio/aac, audio/flac, audio/wav, audio/pcm

Binary audio stream.

Response headers

X-Request-IDstring

Unique request identifier (req_<id>). Include in support tickets.

Response codes

200

Binary audio stream.

400

Invalid request — unsupported voice/format, input too long, or validation error.

Error

401

Unauthorized — missing or invalid API key.

402

Insufficient credits.

403

API key lacks permission for this model.

404

Model not found or not configured for this workspace.

500

Server error.