人们对大型语言模型有两种不同的期待。第一个期待是,我们期待大型语言模型成为解决某一种问题的专材,它可以解决某一种自然语言处理的特定任务。举例来说,我们把大型语言模型做一些调整以后,叫他专门做翻译,你给他什么句子,他都把中文翻成英文,我给他一个句子,他专门给他一段文章,他专门做摘要,看到一段文章作为输入,他就是会把文章变短,这个是让大型语言模型变成专才。那第二个期待是期待大型语言模型变成一个通才,变得无所不能,什么都会
举例来说,你给他一个句子以后,那他因为什么都会。所以如果你只给他一个句子,他不知道他要得到什么样的输出。所以你需要对他用人类可以看得懂的语言直接下指令。你告诉大型语言模型说现在把这句话做翻译,期待它就产生英文的句子。同样的句子,如果你下不同的指令说要做摘要,它就把你输入的文字进行摘要。而这个额外所下的用人类语言来描述的指令,期待今天机器可以看得懂人下的指令。
Speechflow’s ASR API transcribes with a leading accuracy rate in 14 languages, and the list is growing. This makes SpeechFlow an ideal solution for translating audio to text and speech to text.
Using our AI model, audio is transformed into text with proper punctuation, and optimized for reading, resulting in transcriptions that are easy to understand and act upon.
With our simple API design, deploying SpeechFlow is made hassle-free. We support both cloud and on-prem deployment to ensure security, reliability and flexibility.
SpeechFlow can process up to 1 hour of audio file in less than 3 minutes, making it an incredibly efficient solution for businesses and individuals who rely on accurate and timely transcription services.
Billed pay as you go at $0.0002 per second. Have full control and transparency of how much you use and pay for.
First, copy the API KEY ID and API KEY SECRET into the following code. Run the code in the terminal and the taskId will be returned soon.
# for remote file API_KEY_ID="YOUR_API_KEY_ID" API_KEY_SECRET="YOUR_API_KEY_SECRET" # See more lang code: https://docs.speechflow.io/#/?id=ap-lang-list LANG="en" FILE_REMOTE_PATH="https://sf-docs-prod.s3.us-west-1.amazonaws.com/web/sample-audios/EN.wav" curl -H "keyId:${API_KEY_ID}" -H "keySecret:${API_KEY_SECRET}" -X POST -d "lang=${LANG}&remotePath=${FILE_REMOTE_PATH}" \ "https://api.speechflow.io/asr/file/v1/create"
# for local file API_KEY_ID="YOUR_API_KEY_ID" API_KEY_SECRET="YOUR_API_KEY_SECRET" # See more lang code: https://docs.speechflow.io/#/?id=ap-lang-list LANG="en" FILE_LOCAL_PATH="YOUR_FILE_LOCAL_PATH" curl -H "keyId:${API_KEY_ID}" -H "keySecret:${API_KEY_SECRET}" -F "file=@${FILE_LOCAL_PATH}" \ -X POST "https://api.speechflow.io/asr/file/v1/create?lang=${LANG}"
Second, copy the taskId into the following code. Run the code in the terminal and the transcription result will be returned.
API_KEY_ID="YOUR_API_KEY_ID" API_KEY_SECRET="YOUR_API_KEY_SECRET" TASK_ID="THE_TASKID_RETURNED_IN_THE_FIRST_STEP" curl -H "keyId:${API_KEY_ID}" -H "keySecret:${API_KEY_SECRET}" \ "https://api.speechflow.io/asr/file/v1/query?taskId=${TASK_ID}"