fastwhisper

A wraps for fast-whisper with following features:

• a web server
• with following API

1 transcribe

sumits a asr task, and return a dict contains 'task_id' which can use to query task status

2 status

inquery a task status, if 'SUCCEEDED', the result of transcribe will carry out

3 asr

sumbit a asr task and wait it success or failed

Deperndents

fastwhisper will installs all its dependents when install

Download and Install

use follow command to download fastwhisper

git clone https://git.opencomputing.cn/yumoqing/fastwhisper.git

go to the folder 'fastwhisper'

cd fastwhisper

change some config items

vi conf/config.json

modify the following items

	"model_path": "/data/ymq/models/deepdml/faster-whisper-large-v3-turbo-ct2",
	"redis_url": "redis://127.0.0.1:6379",
	"worker_cnt": 1,
	    "port":9925,

finally

bash ./build.sh

./build.sh will 1 make virtual environment in "py3" folder 2 install all packages it needs 3 use "APT" install redis-server, if not on ubuntu, you need to modify build.sh 4 create a system service named "fastwhisper" 5 start fastwhisper servcie

APIs

transcribe

• PATH: /v1/transcribe

• headers we need the following header

{
"Content-Type": "application/json"
}

• data

{
	"audio_file": a.wav or b.mp3
}

• response

{
	"task_id": "13hewndskj83r3"
}

• curl examples

curl -X POST http://127.0.0.1:9925/v1/transcribe \
    -H "Content-TYpe: application/json" \
    -d '{
    "audio_file": "1.wav"
}'

status

• path: /v1/status

• params

{
	"task_id": "task id return from transcribe api"
}

• response

{
    "finished_at": "1762402570.5322073",
    "task_id": "23z2tWN6Z7hlqCgvusigZ",
    "status": "SUCCEEDED",
    "payload": "{\"audio_file\": \"1.wav\"}",
    "started_at": "1762402566.946902",
    "result": {
        "language": "zh",
        "language_probability": 0.99755859375,
        "segments": [
            [
                0.0,
                29.98,
                "优优独播剧场——YoYo Television Series Exclusive"
            ],
            [
                123.78,
                126.66000000000001,
                "产品和研发"
            ],
			......
            [
                126.66000000000001,
                128.22,
                "哦"
            ],
            [
                128.22,
                130.85999999999999,
                "一起把未来憧憬"
            ],
        ]
    },
    "created_at": "1762402566.9158587"
}

response explanation

• status

prosible values are "PENGING":means not starting running, "RUNNING": is running, "FAILED": finished with error, "SUCCEEDED": finished without error

• result

1 'FAILED'

result is the error string

2 'SUCCEEDED'

result is the transcribe result dict keys in result:

I language

identified language

II language_probability

probability of language

III segments it is a list of segment

IV segment start end text

• curl example

url http://127.0.0.1:9925/v1/status?task_id=23z2tWN6Z7hlqCgvusigZ

asr

• path: /v1/asr

• headers we need the following header

{
"Content-Type": "application/json"
}

• data

{
    "audio_file": a.wav or b.mp3
}

• response

{
    "finished_at": "1762402570.5322073",
    "task_id": "23z2tWN6Z7hlqCgvusigZ",
    "status": "SUCCEEDED",
    "payload": "{\"audio_file\": \"1.wav\"}",
    "started_at": "1762402566.946902",
    "result": {
        "language": "zh",
        "language_probability": 0.99755859375,
        "segments": [
            [
                0.0,
                29.98,
                "优优独播剧场——YoYo Television Series Exclusive"
            ],
            [
                123.78,
                126.66000000000001,
                "产品和研发"
            ],
            ......
            [
                126.66000000000001,
                128.22,
                "哦"
            ],
            [
                128.22,
                130.85999999999999,
                "一起把未来憧憬"
            ],
        ]
    },
    "created_at": "1762402566.9158587"
}