text-to-speech
Produce clean, aligned, delivery-ready TTS segments with loudness normalization, boundary cleanup, and export specs for dubbing.
- License 许可
- Apache-2.0
- 最近更新
- 2026-11-14
- Stars
- ★ 0
- Forks
- 1
01 前言 Front Matter
- name
- text-to-speech
- description
- Produce clean, aligned, delivery-ready TTS segments with loudness normalization, boundary cleanup, and export specs for dubbing.
03 SKILL.md 正文
Steps
- Generate TTS audio per segment using an appropriate engine (neural offline or cloud), then confirm the TTS output’s native sample rate.
- Clean up each segment (lightweight, consistent chain):
- High-pass around ~20 Hz (rumble/DC removal)
- Optional low-pass around ~16 kHz (reduce digital fizz)
- Add short fades at boundaries (e.g., ~50 ms) to prevent clicks
- Control duration to fit the timing window:
- If shorter than the window: pad silence
- If longer: apply gentle rate_adjust (small tempo change) or trim carefully
- Loudness workflow (ITU-R BS.1770-4 target behavior):
- Measure loudness with
ebur128for reporting/validation. - Normalize as a final step with
loudnorm(e.g.,I=-23,TP=-1.5,LRA=11). - Do not rely on
loudnormoptions that may not exist in the environment (notablymeasure_only=1on some FFmpeg builds); useebur128for measurement andloudnormfor normalization.
- Measure loudness with
- Export delivery audio specs:
- Ensure output is 48,000 Hz, mono, and a suitable PCM WAV for segment deliverables (e.g.,
pcm_s16le).
- Ensure output is 48,000 Hz, mono, and a suitable PCM WAV for segment deliverables (e.g.,
Expected Result
Each TTS segment is clean, click-free, window-aligned with controlled duration (rate_adjust/pad_silence/trim), normalized to a consistent loudness target, and exported as 48 kHz mono audio suitable for muxing into the final dubbed MP4.
04 同仓库更多 Skills
video-frame-extraction
Extract frames from video files and save them as images using OpenCV
openai-vision
Analyze images and multi-frame sequences using OpenAI GPT vision models
image-ocr
Extract text content from images using Tesseract OCR via Python, including structured field extraction from receipts and forms
video-frame-extraction
Extract frames from video files and save them as images using OpenCV
openai-vision
Analyze images and multi-frame sequences using OpenAI GPT vision models
image-ocr
Extract text content from images using Tesseract OCR via Python, including structured field extraction from receipts and forms