SenseAudio Language Tutor

Create interactive language-learning audio with official SenseAudio TTS endpoints and parameters.

What This Skill Does

• Generate pronunciation examples in supported voices
• Create bilingual vocabulary and sentence practice audio
• Produce slowed-speed listening drills for learners
• Build short dialogue exercises with repetition pauses
• Export lesson audio files and companion study notes

Credential and Dependency Rules

• Read the API key from SENSEAUDIO_API_KEY.
• Send auth only as Authorization: Bearer \x3CAPI_KEY>.
• Do not place API keys in query parameters, logs, or saved examples.
• If Python helpers are used, this skill expects python3, requests, and pydub.
• pydub may also require a local audio backend such as ffmpeg; if unavailable, prefer writing individual audio files instead of merging them.

Official TTS Constraints

Use the official SenseAudio TTS rules summarized below:

• HTTP endpoint: POST https://api.senseaudio.cn/v1/t2a_v2
• Model: SenseAudio-TTS-1.0
• Max text length: 10000 characters
• voice_setting.voice_id is required
• voice_setting.speed range: 0.5-2.0
• Optional audio format values: mp3, wav, pcm, flac
• Optional sample rates: 8000, 16000, 22050, 24000, 32000, 44100
• Optional MP3 bitrates: 32000, 64000, 128000, 256000
• Optional channels: 1 or 2

Recommended Workflow

1. Prepare lesson content:

• Split vocabulary, example sentences, and dialogues into short chunks.
• Keep each API call comfortably below the 10000 character limit.

2. Build minimal TTS requests:

• Send model, text, stream, and voice_setting.voice_id.
• Add speed, pitch, vol, and audio_setting only when needed.

3. Decode and save audio safely:

• HTTP responses return hex-encoded audio in data.audio; decode before saving.
• Keep filenames deterministic and avoid exposing secrets in paths or logs.

4. Compose lessons carefully:

• If pydub and an audio backend are available, merge clips and insert silence.
• Otherwise, emit per-word or per-sentence clips and a manifest/Markdown study guide.

5. Handle failures and traceability:

• Check HTTP status and provider error payloads before decoding audio.
• Record trace_id only for troubleshooting and avoid showing it unless needed.

Minimal Helper

import binascii
import os

import requests

API_KEY = os.environ["SENSEAUDIO_API_KEY"]
API_URL = "https://api.senseaudio.cn/v1/t2a_v2"


def generate_tts(text, voice_id="male_0004_a", speed=1.0, stream=False):
    payload = {
        "model": "SenseAudio-TTS-1.0",
        "text": text,
        "stream": stream,
        "voice_setting": {
            "voice_id": voice_id,
            "speed": speed,
        },
        "audio_setting": {
            "format": "mp3",
            "sample_rate": 32000,
            "bitrate": 128000,
            "channel": 2,
        },
    }
    response = requests.post(
        API_URL,
        headers={
            "Authorization": f"Bearer {API_KEY}",
            "Content-Type": "application/json",
        },
        json=payload,
        timeout=60,
    )
    response.raise_for_status()
    data = response.json()
    audio_hex = data["data"]["audio"]
    return binascii.unhexlify(audio_hex), data.get("trace_id")

Patterns

Vocabulary Drill

• Generate one clip for the target word
• Generate one clip for an example sentence
• Optionally generate a slower clip at speed=0.8
• Save clips separately or merge with pauses

Bilingual Lesson

• Alternate source phrase and translated phrase
• Use short pauses (1000-2000ms) between clips
• Consider different voice_id values for source and translation when helpful

Dialogue Practice

• Create one clip per line of dialogue
• Insert repetition pauses after each line
• Prefer shorter turns for easier debugging and regeneration

Output Options

• Individual MP3 clips for words, sentences, or dialogue turns
• Merged lesson audio if local audio tooling is available
• Markdown study guide with transcript, translation, and file manifest

Safety Notes

• Do not hardcode credentials.
• Do not claim unsupported language-selection parameters for TTS unless the official docs add them.
• Avoid assuming raw bytes can be passed directly to pydub.AudioSegment; decode and load through a supported container format.