README.md

# whisper-timestamped

Multilingual Automatic Speech Recognition with word-level timestamps and confidence.

* [Description](#description)
   * [Notes on other approaches](#notes-on-other-approaches)
* [Installation](#installation)
   * [First installation](#first-installation)
      * [Additional packages that might be needed](#additional-packages-that-might-be-needed)
      * [Docker](#docker)
   * [Light installation for CPU](#light-installation-for-cpu)
   * [Upgrade to the latest version](#upgrade-to-the-latest-version)
* [Usage](#usage)
   * [Python](#python)
   * [Command line](#command-line)
   * [Plotting word alignment](#plotting-word-alignment)
   * [Example output](#example-output)
   * [Options that may improve results](#options-that-may-improve-results)
      * [Accurate Whisper transcription](#accurate-whisper-transcription)
      * [Running Voice Activity Detection (VAD) before sending to Whisper](#running-voice-activity-detection-vad-before-sending-to-whisper)
      * [Detecting disfluencies](#detecting-disfluencies)
* [Acknowlegment](#acknowlegment)
* [Citations](#citations)

## Description

[Whisper](https://openai.com/blog/whisper/) is a set of multi-lingual, robust speech recognition models trained by OpenAI that achieve state-of-the-art results in many languages. Whisper models were trained to predict approximate timestamps on speech segments (most of the time with 1-second accuracy), but they cannot originally predict word timestamps. This repository proposes an implementation to **predict word timestamps and provide a more accurate estimation of speech segments when transcribing with Whisper models**.
Besides, a confidence score is assigned to each word and each segment.

The approach is based on Dynamic Time Warping (DTW) applied to cross-attention weights, as demonstrated by [this notebook by Jong Wook Kim](https://github.com/openai/whisper/blob/f82bc59f5ea234d4b97fb2860842ed38519f7e65/notebooks/Multilingual_ASR.ipynb). There are some additions to this notebook:
* The start/end estimation is more accurate.
* Confidence scores are assigned to each word.
* **If possible (without beam search...)**, no additional inference steps are required to predict word timestamps (word alignment is done on the fly after each speech segment is decoded).
* Special care has been taken regarding memory usage: `whisper-timestamped` is able to process long files with little additional memory compared to the regular use of the Whisper model.

`whisper-timestamped` is an extension of the [`openai-whisper`](https://pypi.org/project/whisper-openai/) Python package and is meant to be compatible with any version of `openai-whisper`.
It provides more efficient/accurate word timestamps, along with those additional features:
* Voice Activity Detection (VAD) can be run before applying Whisper model,
  to avoid hallucinations due to errors in the training data (for instance, predicting "Thanks you for watching!" on pure silence).
  Several VAD methods are available: silero (default), auditok, auditok:v3.1
* When the language is not specified, the language probabilities are provided among the outputs.

### Notes on other approaches

An alternative relevant approach to recovering word-level timestamps involves using wav2vec models that predict characters, as successfully implemented in [whisperX](https://github.com/m-bain/whisperX). However, these approaches have several drawbacks that are not present in approaches based on cross-attention weights such as `whisper_timestamped`. These drawbacks include:
* The need to find one wav2vec model per language to support, which does not scale well with the multi-lingual capabilities of Whisper.
* The need to handle (at least) one additional neural network (wav2vec model), which consumes memory.
* The need to normalize characters in Whisper transcription to match the character set of the wav2vec model. This involves awkward language-dependent conversions, such as converting numbers to words ("2" -> "two"), symbols to words ("%" -> "percent", "€" -> "euro(s)")...
* The lack of robustness around speech disfluencies (fillers, hesitations, repeated words...) that are usually removed by Whisper.

An alternative approach that does not require an additional model is to look at the probabilities of timestamp tokens estimated by the Whisper model after each (sub)word token is predicted. This was implemented, for instance, in whisper.cpp and stable-ts. However, this approach lacks robustness because Whisper models have not been trained to output meaningful timestamps after each word. Whisper models tend to predict timestamps only after a certain number of words have been predicted (typically at the end of a sentence), and the probability distribution of timestamps outside this condition may be inaccurate. In practice, these methods can produce results that are totally out-of-sync on some periods of time (we observed this especially when there is jingle music). Also, the timestamp precision of Whisper models tends to be rounded to 1 second (as in many video subtitles), which is too inaccurate for words, and reaching better accuracy is tricky.

## Installation

### First installation

Requirements:
* `python3` (version higher or equal to 3.7, at least 3.9 is recommended)
* `ffmpeg` (see instructions for installation on the [whisper repository](https://github.com/openai/whisper))

You can install `whisper-timestamped` either by using pip:
```bash
pip3 install whisper-timestamped
```

or by cloning this repository and running installation:
```bash
git clone https://github.com/linto-ai/whisper-timestamped
cd whisper-timestamped/
python3 setup.py install
```

#### Additional packages that might be needed

If you want to plot alignment between audio timestamps and words (as in [this section](#plotting-word-alignment)), you also need matplotlib:
```bash
pip3 install matplotlib
```

If you want to use VAD option (Voice Activity Detection before running Whisper model), you also need torchaudio and onnxruntime:
```bash
pip3 install onnxruntime torchaudio
```

If you want to use finetuned Whisper models from the Hugging Face Hub, you also need transformers:
```bash
pip3 install transformers
```

#### Docker

A docker image of about 9GB can be built using:
```bash
git clone https://github.com/linto-ai/whisper-timestamped
cd whisper-timestamped/
docker build -t whisper_timestamped:latest .
```

### Light installation for CPU

If you don't have a GPU (or don't want to use it), then you don't need to install the CUDA dependencies. You should then just install a light version of torch **before** installing whisper-timestamped, for instance as follows:
```bash
pip3 install \
     torch==1.13.1+cpu \
     torchaudio==0.13.1+cpu \
     -f https://download.pytorch.org/whl/torch_stable.html
```

A specific docker image of about 3.5GB can also be built using:
```bash
git clone https://github.com/linto-ai/whisper-timestamped
cd whisper-timestamped/
docker build -t whisper_timestamped_cpu:latest -f Dockerfile.cpu .
```

### Upgrade to the latest version

When using pip, the library can be updated to the latest version using:
```
pip3 install --upgrade --no-deps --force-reinstall git+https://github.com/linto-ai/whisper-timestamped
```

A specific version of `openai-whisper` can be used by running, for example:
```bash
pip3 install openai-whisper==20230124
```

## Usage

### Python

In Python, you can use the function `whisper_timestamped.transcribe()`, which is similar to the function `whisper.transcribe()`:
```python
import whisper_timestamped
help(whisper_timestamped.transcribe)
```
The main difference with `whisper.transcribe()` is that the output will include a key `"words"` for all segments, with the word start and end position. Note that the word will include punctuation. See the example [below](#example-output).

Besides, the default decoding options are different to favour efficient decoding (greedy decoding instead of beam search, and no temperature sampling fallback). To have same default as in `whisper`, use ```beam_size=5, best_of=5, temperature=(0.0, 0.2, 0.4, 0.6, 0.8, 1.0)```.

There are also additional options related to word alignement.

In general, if you import `whisper_timestamped` instead of `whisper` in your Python script and use `transcribe(model, ...)` instead of `model.transcribe(...)`, it should do the job:
```
import whisper_timestamped as whisper

audio = whisper.load_audio("AUDIO.wav")

model = whisper.load_model("tiny", device="cpu")

result = whisper.transcribe(model, audio, language="fr")

import json
print(json.dumps(result, indent = 2, ensure_ascii = False))
```

Note that you can use a finetuned Whisper model from HuggingFace or a local folder by using the `load_model` method of `whisper_timestamped`. For instance, if you want to use [whisper-large-v2-nob](https://huggingface.co/NbAiLab/whisper-large-v2-nob), you can simply do the following:
```
import whisper_timestamped as whisper

model = whisper.load_model("NbAiLab/whisper-large-v2-nob", device="cpu")

# ...
```

### Command line

You can also use `whisper_timestamped` on the command line, similarly to `whisper`. See help with:
```bash
whisper_timestamped --help
```

The main differences with `whisper` CLI are:
* Output files:
  * The output JSON contains word timestamps and confidence scores. See example [below](#example-output).
  * There is an additional CSV output format.
  * For SRT, VTT, TSV formats, there will be additional files saved with word timestamps.
* Some default options are different:
  * By default, no output folder is set: Use `--output_dir .` for Whisper default.
  * By default, there is no verbose: Use `--verbose True` for Whisper default.
  * By default, beam search decoding and temperature sampling fallback are disabled, to favour an efficient decoding.
    To set the same as Whisper default, you can use `--accurate` (which is an alias for ```--beam_size 5 --temperature_increment_on_fallback 0.2 --best_of 5```).
* There are some additional specific options:
  <!-- * `--efficient` to use a faster greedy decoding (without beam search neither several sampling at each step),
  which enables a special path where word timestamps are computed on the fly (no need to run inference twice).
  Note that transcription results might be significantly worse on challenging audios with this option. -->
  * `--compute_confidence` to enable/disable the computation of confidence scores for each word.
  * `--punctuations_with_words` to decide whether punctuation marks should be included or not with preceding words.

An example command to process several files using the `tiny` model and output the results in the current folder, as would be done by default with whisper, is as follows:
```
whisper_timestamped audio1.flac audio2.mp3 audio3.wav --model tiny --output_dir .
```

Note that you can use a fine-tuned Whisper model from HuggingFace or a local folder. For instance, if you want to use the [whisper-large-v2-nob](https://huggingface.co/NbAiLab/whisper-large-v2-nob) model, you can simply do the following:
```
whisper_timestamped --model NbAiLab/whisper-large-v2-nob <...>
```

### Plot of word alignment

Note that you can use the `plot_word_alignment` option of the `whisper_timestamped.transcribe()` Python function or the `--plot` option of the `whisper_timestamped` CLI to see the word alignment for each segment.

![Example alignement](figs/example_alignement_plot.png)

* The upper plot represents the transformation of cross-attention weights used for alignment with Dynamic Time Warping. The abscissa represents time, and the ordinate represents the predicted tokens, with special timestamp tokens at the beginning and end, and (sub)words and punctuation in the middle.
* The lower plot is an MFCC representation of the input signal (features used by Whisper, based on Mel-frequency cepstrum).
* The vertical dotted red lines show where the word boundaries are found (with punctuation marks "glued" to the previous word).

### Example output

Here is an example output of the `whisper_timestamped.transcribe()` function, which can be viewed by using the CLI:
```bash
whisper_timestamped AUDIO_FILE.wav --model tiny --language fr
```
```json
{
  "text": " Bonjour! Est-ce que vous allez bien?",
  "segments": [
    {
      "id": 0,
      "seek": 0,
      "start": 0.5,
      "end": 1.2,
      "text": " Bonjour!",
      "tokens": [ 25431, 2298 ],
      "temperature": 0.0,
      "avg_logprob": -0.6674491882324218,
      "compression_ratio": 0.8181818181818182,
      "no_speech_prob": 0.10241222381591797,
      "confidence": 0.51,
      "words": [
        {
          "text": "Bonjour!",
          "start": 0.5,
          "end": 1.2,
          "confidence": 0.51
        }
      ]
    },
    {
      "id": 1,
      "seek": 200,
      "start": 2.02,
      "end": 4.48,
      "text": " Est-ce que vous allez bien?",
      "tokens": [ 50364, 4410, 12, 384, 631, 2630, 18146, 3610, 2506, 50464 ],
      "temperature": 0.0,
      "avg_logprob": -0.43492694334550336,
      "compression_ratio": 0.7714285714285715,
      "no_speech_prob": 0.06502953916788101,
      "confidence": 0.595,
      "words": [
        {
          "text": "Est-ce",
          "start": 2.02,
          "end": 3.78,
          "confidence": 0.441
        },
        {
          "text": "que",
          "start": 3.78,
          "end": 3.84,
          "confidence": 0.948
        },
        {
          "text": "vous",
          "start": 3.84,
          "end": 4.0,
          "confidence": 0.935
        },
        {
          "text": "allez",
          "start": 4.0,
          "end": 4.14,
          "confidence": 0.347
        },
        {
          "text": "bien?",
          "start": 4.14,
          "end": 4.48,
          "confidence": 0.998
        }
      ]
    }
  ],
  "language": "fr"
}
```
If the language is not specified (e.g. without option `--language fr` in the CLI) you will find an additional key with the language probabilities:
```json
{
  ...
  "language": "fr",
  "language_probs": {
    "en": 0.027954353019595146,
    "zh": 0.02743500843644142,
    ...
    "fr": 0.9196318984031677,
    ...
    "su": 3.0119704064190955e-08,
    "yue": 2.2565967810805887e-05
  }
}
```

### Options that may improve results

Here are some options that are not enabled by default but might improve results.

#### Accurate Whisper transcription

As mentioned earlier, some decoding options are disabled by default to offer better efficiency. However, this can impact the quality of the transcription. To run with the options that have the best chance of providing a good transcription, use the following options.
* In Python:
```python
results = whisper_timestamped.transcribe(model, audio, beam_size=5, best_of=5, temperature=(0.0, 0.2, 0.4, 0.6, 0.8, 1.0), ...)
```
* On the command line:
```bash
whisper_timestamped --accurate ...
```

#### Running Voice Activity Detection (VAD) before sending to Whisper

Whisper models can "hallucinate" text when given a segment without speech. This can be avoided by running VAD and gluing speech segments together before transcribing with the Whisper model. This is possible with `whisper-timestamped`.
* In Python:
```python
results = whisper_timestamped.transcribe(model, audio, vad=True, ...)
```
* On the command line:
```bash
whisper_timestamped --vad True ...
```

By default, the VAD method used is [silero](https://github.com/snakers4/silero-vad).
But other methods are available, such as earlier versions of silero, or [auditok](https://github.com/amsehili/auditok).
Those methods were introduced because latest versions of silero VAD can have a lot of false alarms on some audios (speech detected on silence).
* In Python:
```python
results = whisper_timestamped.transcribe(model, audio, vad="silero:v3.1", ...)
results = whisper_timestamped.transcribe(model, audio, vad="auditok", ...)
```
* On the command line:
```bash
whisper_timestamped --vad silero:v3.1 ...
whisper_timestamped --vad auditok ...
```

In order to watch the VAD results, you can use the `--plot` option of the `whisper_timestamped` CLI,
or the `plot_word_alignment` option of the `whisper_timestamped.transcribe()` Python function.
It will show the VAD results on the input audio signal as following (x-axis is time in seconds):
| **vad="silero:v4.0"** | **vad="silero:v3.1"** | **vad="auditok"** |
| :---: | :---: | :---: |
| ![Example VAD](figs/VAD_silero_v4.0.png) | ![Example VAD](figs/VAD_silero_v3.1.png)  | ![Example VAD](figs/VAD_auditok.png) |

#### Detecting disfluencies

Whisper models tend to remove speech disfluencies (filler words, hesitations, repetitions, etc.). Without precautions, the disfluencies that are not transcribed will affect the timestamp of the following word: the timestamp of the beginning of the word will actually be the timestamp of the beginning of the disfluencies. `whisper-timestamped` can have some heuristics to avoid this.
* In Python:
```python
results = whisper_timestamped.transcribe(model, audio, detect_disfluencies=True, ...)
```
* On the command line:
```bash
whisper_timestamped --detect_disfluencies True ...
```
**Important:** Note that when using these options, possible disfluencies will appear in the transcription as a special "`[*]`" word.


## Acknowlegment
* [whisper](https://github.com/openai/whisper): Whisper speech recognition (License MIT).
* [dtw-python](https://pypi.org/project/dtw-python): Dynamic Time Warping (License GPL v3).

## Citations
If you use this in your research, please cite the repo:

```bibtex
@misc{lintoai2023whispertimestamped,
  title={whisper-timestamped},
  author={Louradour, J{\'e}r{\^o}me},
  journal={GitHub repository},
  year={2023},
  publisher={GitHub},
  howpublished = {\url{https://github.com/linto-ai/whisper-timestamped}}
}
```

as well as the OpenAI Whisper paper:

```bibtex
@article{radford2022robust,
  title={Robust speech recognition via large-scale weak supervision},
  author={Radford, Alec and Kim, Jong Wook and Xu, Tao and Brockman, Greg and McLeavey, Christine and Sutskever, Ilya},
  journal={arXiv preprint arXiv:2212.04356},
  year={2022}
}
```

and this paper for Dynamic-Time-Warping:

```bibtex
@article{JSSv031i07,
  title={Computing and Visualizing Dynamic Time Warping Alignments in R: The dtw Package},
  author={Giorgino, Toni},
  journal={Journal of Statistical Software},
  year={2009},
  volume={31},
  number={7},
  doi={10.18637/jss.v031.i07}
}
```