Announcement

Collapse
No announcement yet.

Speech to Text Software

Collapse
X
 
  • Filter
  • Time
  • Show
Clear All
new posts

    Speech to Text Software

    Hi

    I have about 45 hour to hour and a half event speeches that I need to get the audio speech from the video converted into text to be searchable. I notice that Premiere has thrown in the towel and I am looking for something that works. A search keeps turning up Nuance's Dragon Naturally Speaking. Does anyone have experience with it for this purpose. Or some other solution. Thanks in advance.

    #2
    I have a friend with serious carpal-tunnel issues who's been using Dragon Naturally Speaking for all of his text input for almost a decade. It's not 100 percent perfect but it's considered to be the best software on the market for this.

    Comment


      #3
      Hi
      Thanks for the response. I just emailed Nuance with the question to see what they say. I wonder if what works for simple commands or can be repeated in real time as needed fails for transcribing from a video? Anyone try this?

      Don't want to hijack my own thread! But I noticed the one product they say is 508 certified and for the disabled is their most expensive at twice the price of their next most expensive personal product. Way to go Nuance. I hope I am missing something there. I really do.

      Comment


        #4
        The engine in the Nuance software has continued to improve since I started working with it about 12 years ago. There are a lot of variables that go into the usefulness of Dragon to achieve your objective such as the quality of the audio and the characteristics of the speech being recognized. There are settings in Dragon to improve grammatical interpretation to lessen errors from homonyms and the like. Ultimately it becomes a measure of the accuracy and labor requirement for correcting errors vs. the cost of a transcriptionist, as well as your time constraints. When I have trained Dragon to recognize my speech I still have some corrections to make to dictation content.

        Comment


          #5
          It has been a while since I used them, but I found that the Nuance products are excellent, once they were "trained." That is, you need to go through a series of phrases to teach them how you speak. Might not work for your purposes.

          Not sure it would do an hour-and-a-half, but my iPhone does a fantastic job of text to speech.

          Jim

          Comment


            #6
            The progress in the past couple of years has to do with speaker independent voice recognition, so less training required. Years ago I spent a lot of time doing research on OCR for a project. It turned out we had a typist on staff who was much cheaper and more accurate than the OCR software would have been and project completed in days rather than months.

            Comment


              #7
              Hi
              Thanks for the great responses. I also had a copy of Dragon many years ago. Training the software was important to success. I am not sure your can do that with a video. Unless someone has recently tried Dragon with a video and can report results I guess my only option is to buy and try. Which version might be the next question.

              Comment


                #8
                If you have the budget, there are transcription services out there too
                sigpic

                Independent Filmmaker
                BMD URSA Mini 4K/Avid Media Composer/NukeX/Blender/Mixcraft/ProTools/Resolve Studio

                Feature Films
                Wulf - 2008 | Leap - 2010 | Leap: Rise of the Beast - 2011 | Surviving The Wild - 2020

                Comment


                  #9
                  Hi
                  Thanks for the response. I should have mentioned that I am looking into this on behalf of a small historical society. They have been videoing historical presentations going back to 1999 and are now entering them in a "OMEKA" database. It would be great to make the presentation searchable in the database. They don't have the volunteer manpower to transcribe or the money to hire out. Even an less than perfect text could be very useful.

                  Comment


                    #10
                    I have used Nuance Dragon for many years, it still has issues, even when a $350 mic is used and it has been trained to recognize my voice. I'd suggest first step is to try a recent ipad, built in voice recognition, I believe powered by Nuance. Just hold it up to a loudspeaker. If this doesn't work try Dragon. You'll probably need the premium version, the home versions has some limitations. I'd suggest at least i5 4GB Ram. Laptops often have noisy soundcards, USB mic is usually preferred, requires premiere version. Dragon until recently required the speaker to train it by reading certain text for 10 minutes, the last professional version no longer mandates this (not sure about premium version). I hope each of your presentations consist of 1 person speaking to permit Dragon to learn their voice, you'd have to set them up as seperate profiles in the software, I think again requires premium version. There is one other voice recognition company that I am aware of, M Modal Fluency. They are cloud based, but do speech recognition via phone line, so less than ideal audio input. I suspect however that they only deal with large customers (they are doing a trial at a local hospital presently)

                    Comment


                      #11
                      Hi
                      THANKS! for the response. Really good information. I will try ipad. Great place to start. The historical society does have a video editing station with a great PC and a RME Babyface sound interface. Also decent mics, mostly Audio Technica. I was hoping to route the sound from the video directly into the Dragon software without going the speaker out and mic in routine.

                      Most of the presentations are one main speaker but also have introductions and some questions. I may have to just have the society buy it and see. ipad first though! I will try to post back with results for the ipad test.

                      Comment


                        #12
                        Any feedback on this? I have a few hour long lectures that need to be transcribed for editing purposes. The transcription doesn't have to be 100% perfect, just enough so I can do a paper edit of the dialog before editing the actual video.

                        Again, I'm not interested in talking into my computer, I just want to turn the audio from my recordings into text. Is Dragon the best software for this?

                        Comment


                          #13
                          I got an interesting flyer in the mail from Nuance Dragon manufacturer that addresses this directly. They state "In answer to one of our biggest customer requests, Dragon Professional Individual will now accurately transcribe any single speakerís voice from pre-recorded audio files or podcasts, requiring just a 90-second audio clip to create the necessary profile. Now itís a breeze to create a written transcript of that keynote speech which impressed you so much. " The website is here: http://shop.nuance.com/store?SiteID=...Page=PostalDPI , they seem to be having a special til the end of September, $99 for Dragon professional (AFAIK there is a cheaper Dragon basic or home version which may not work for this). Sound to me like Nuance have done exactly what the OP and the @nickbeef are looking for..

                          Comment


                            #14
                            I hope you'll report back your findings. My guess is you'll achieve about 70% accuracy if you have clear audio.

                            Comment


                              #15
                              Perfect. And with a 30-day trial, it will be great to test how much it works/doesn't work.

                              Comment

                              Working...
                              X