Volcengine ATA Subtitle (自动打轴)

Generate subtitles with automatic time alignment using Volcengine's ATA (Automatic Time Alignment) API.

Prerequisites

Set the following environment variables or create a config file:

Option A: Environment Variables

export VOLC_ATA_APP_ID="your-app-id"
export VOLC_ATA_TOKEN="your-access-token"
export VOLC_ATA_API_BASE="https://openspeech.bytedance.com"

Option B: Config File

Create ~/.volcengine_ata.conf:

[credentials]
appid = your-app-id
access_token = your-access-token
secret_key = your-secret-key

[api]
base_url = https://openspeech.bytedance.com
submit_path = /api/v1/vc/ata/submit
query_path = /api/v1/vc/ata/query

Execution (Python CLI Tool)

A Python CLI tool is provided at ~/.openclaw/workspace/skills/volcengine-ata-subtitle/volc_ata.py.

Quick Examples

# Basic usage: audio + text → SRT subtitle
python3 ~/.openclaw/workspace/skills/volcengine-ata-subtitle/volc_ata.py \
  --audio storage/audio.wav \
  --text storage/subtitle.txt \
  --output storage/subtitles/final.srt

# Specify output format (srt or ass)
python3 ~/.openclaw/workspace/skills/volcengine-ata-subtitle/volc_ata.py \
  --audio storage/audio.wav \
  --text storage/subtitle.txt \
  --output storage/subtitles/final.ass \
  --format ass

Input Requirements

Audio File

Format: WAV (PCM)
Sample Rate: 16000 Hz (16kHz)
Channels: 1 (mono)
Encoding: 16-bit PCM (pcm_s16le)

Extract from video:

ffmpeg -i input.mp4 -vn -acodec pcm_s16le -ar 16000 -ac 1 audio.wav

Text File

Format: Plain text (UTF-8)
Structure: One sentence per line
No punctuation: ATA will handle automatically
No timestamps: Pure text only

Example:

主人闹钟没响睡过头了
我们俩轮流用鼻子拱他脸
他以为地震了抱着枕头就跑

Output Formats

SRT (SubRip)

1
00:00:00,000 --> 00:00:02,500
第一句字幕

2
00:00:02,500 --> 00:00:05,000
第二句字幕

ASS (Advanced Substation Alpha)

[Script Info]
Title: ATA Subtitles
ScriptType: v4.00+

[Events]
Format: Layer, Start, End, Style, Name, MarginL, MarginR, MarginV, Effect, Text
Dialogue: 0,0:00:00.00,0:00:02.50,Default,,0,0,0,,第一句字幕

Rules

Always check that credentials are configured before making API calls.
Audio must be 16kHz mono PCM - convert if necessary with ffmpeg.
Text should be plain - no timestamps, no punctuation.
Default format: SRT (most compatible).
Handle errors gracefully - display clear error messages.

Troubleshooting

Invalid Sample Rate

Error: Invalid sample rate, expected 16000Hz

Fix:

ffmpeg -i input.mp4 -ar 16000 -ac 1 audio.wav

Authorization Failed

Error: Authorization failed

Fix: Check token format. Should be Bearer; {token} (with semicolon).

volcengine-ata-subtitle

Safety Notice

Copy this and send it to your AI assistant to learn