What is forced alignment? It is the matching of a known text to an audio recording in order to find when each word is spoken. The words are given in advance, so the system only decides where they fall in time, and it does not need to work out what was said.
Transcribe a recording
Send your voice message, audio or video to @potok_now_bot on Telegram and the text comes back in the chat.
Open @potok_now_bot in Telegram
Files up to 2 GB · export to TXT, MD, DOCX, PDF, SRT and VTT · speaker labels and timestamps · about 30 languages
Key facts
- The input is both audio and its exact text, such as a script, a book chapter or a corrected transcript.
- The output is a start and end time for every word, which is finer than the line-level times in a subtitle file.
- If the text differs from what was actually said, the timings near the mismatch become unreliable.
- Typical uses are karaoke-style highlighting, subtitle timing and finding a sentence inside a long recording.
What it looks like
input: audio = reading.wav
text = "the harbour was quiet"
output: [
{"word": "the", "start": 0.42, "end": 0.55},
{"word": "harbour","start": 0.55, "end": 1.02},
{"word": "was", "start": 1.02, "end": 1.21},
{"word": "quiet", "start": 1.21, "end": 1.80}
]
More
See the full list of languages, the tools and the glossary.
Frequently asked questions
How is forced alignment different from transcription?
Transcription finds the words from sound. Forced alignment starts with the words and finds only their positions in the audio, so errors come from timing, not from recognition.
What do I need to run it?
You need the recording and a text that matches it closely. The cleaner the match between speech and text, the better the timings.
What happens if the text has mistakes?
Words that were not spoken, or are missing from the text, confuse the alignment around them. Correct the transcript first, then align it.
Why would I want word-level timing?
It lets you jump to a spoken word, highlight text during playback, and cut audio at exact words rather than at the start of a line.
Can I get a timed transcript from the Potok bot?
Yes. Timestamps can be switched on, and you can export SRT or VTT. Check the timing against playback, as long pauses and noise can shift it.
Sources
Each source was opened by hand on the date shown. Pages change, so open the source if a detail matters to you.
- Forced alignment (Wikipedia), Wikipedia · checked
- Speech recognition (Wikipedia), Wikipedia · checked
Try it on your own recording
Send your voice message, audio or video to @potok_now_bot on Telegram and the text comes back in the chat.
Open @potok_now_bot in Telegram