Audio Speech (TTS)

POST /audio/speech matches the OpenAI Text-to-Speech spec — the official openai SDKs work unchanged with just a base URL swap. Accepts model, input, and an optional voice; returns raw audio bytes (MP3, 32 kHz, 128 kbps, mono).

Quickstart

#60a5fa]">from openai import OpenAI

client = OpenAI(
    api_key=#60a5fa]">class="text-emerald-400">"tl-xxxxxxxxxxxxxxxxxxxxxxxx",
    base_url=#60a5fa]">class="text-emerald-400">"https://api.thalam.ai/v1",
)

response = client.audio.speech.create(
    model=#60a5fa]">class="text-emerald-400">"minimax/minimax-speech-2.8-hd",
    input=#60a5fa]">class="text-emerald-400">"Welcome to Thalam, your unified gateway to AI models.",
    voice=#60a5fa]">class="text-emerald-400">"English_Graceful_Lady",
)

# response.content is the raw audio bytes (mp3 by default).
with open(#60a5fa]">class="text-emerald-400">"out.mp3", class="text-emerald-400">"wb") as f:
    f.write(response.content)

#60a5fa]">import OpenAI from class="text-emerald-400">"openai";
#60a5fa]">import fs from class="text-emerald-400">"node:fs";

#60a5fa]">const client = new OpenAI({
  apiKey: process.env.THALAM_KEY,
  baseURL: #60a5fa]">class="text-emerald-400">"https://api.thalam.ai/v1",
});

#60a5fa]">const response = await client.audio.speech.create({
  model: #60a5fa]">class="text-emerald-400">"minimax/minimax-speech-2.8-hd",
  input: #60a5fa]">class="text-emerald-400">"Welcome to Thalam, your unified gateway to AI models.",
  voice: #60a5fa]">class="text-emerald-400">"English_Graceful_Lady",
});

#60a5fa]">const buffer = Buffer.from(await response.arrayBuffer());
fs.writeFileSync(#60a5fa]">class="text-emerald-400">"out.mp3", buffer);

curl https://api.thalam.ai/v1/audio/speech \
  -H #60a5fa]">class="text-emerald-400">"Authorization: Bearer tl-xxxxxxxxxxxxxxxxxxxxxxxx" \
  -H #60a5fa]">class="text-emerald-400">"Content-Type: application/json" \
  -d '{
    #60a5fa]">class="text-emerald-400">"model": class="text-emerald-400">"minimax/minimax-speech-2.8-hd",
    #60a5fa]">class="text-emerald-400">"input": class="text-emerald-400">"Welcome to Thalam.",
    #60a5fa]">class="text-emerald-400">"voice": class="text-emerald-400">"English_Graceful_Lady"
  }' \
  --output out.mp3

Request body

Not yet supported. The OpenAI spec also defines response_format (mp3 / opus / aac / flac / wav / pcm) and speed (0.25 – 4.0). Our gateway accepts these fields but ignores them today — audio is always returned as MP3 at 1.0× speed. MiniMax-specific upstream params (pitch, volume, emotion, language_boost, sample_rate, bitrate) are also silently dropped. Wiring these through is on the roadmap — let us know if you need any of them sooner.

Response

The response body is the raw audio file. No JSON envelope.

Streaming: stream: true is not currently supported on /audio/speech; the gateway waits for the upstream to finish and returns the full file in one response. For typical sentence-length inputs this is ~1–4 seconds end-to-end.

Available models

Live rates are also on the pricing page.

MiniMax voice catalog

All three MiniMax models share the same voice IDs. Pass any of these as the voice field, grouped here by language:

The English / Chinese / Japanese voices are listed in MiniMax's official voice catalog. Custom cloned voice IDs from your MiniMax account also work — the gateway forwards voice through unchanged.

Common errors