5 Matching Annotations
  1. Nov 2023
    1. The best way to use Descript is to edit the waveform with the text as a guide. You really cant edit straight in the text as the results are bad.

      Right on! Going with the text only is asking for trouble. The waveform is still irreplaceable; the words are a guide.

    2. I'm a big fan. Here are a few videos that show my workflow:https://youtu.be/N41v30oLM2ohttps://www.youtube.com/watch?v=XGAz7q_ralkI figure it's actually saved me a ton of time and $, because my time spent editing has gone way down.

      Good argument for it. If one does enough podcasting or vlogging, the $24/month is nothing for all that it can do.

    3. I'm using it currently. I'd say its a mixed bag, but mostly positive. I Um and Err ALOT and wanted a program to help with that. After using it for 6 weeks, I think it works but takes time to utilize properly. The best course of action long term is likely to just learn better speaking habits.Its VERY easy to remove filler words, which is mainly what I use it for. At times it will cut out another word next to the filler word, so you still have to listen to the edit and make sure everything is good. Its not a click once and done type of thing.The transcription itself is pretty good considering I have not provided it with any voice samples. Editing the text and seeing it reflected in the video edit is super cool. It has trouble with words that are unique or proper nouns, but that is to be expected.There is a feature that can recreate your voice - because you need to provide voice samples for that, I have not tried this and therefore cannot speak to how well it works. Again I suspect it would have problems translating proper nouns.Overall, I think it generally does what it advertises well, but people expecting a sliver bullet will be disappointed. My recommendation would be to buy it for a few months first, then sign up for an annual plan if you like it. Its worth a $50 to see if helps you enough to justify the cost.

      Good suggestion. I feel it's going to be a mixed bag too. The AI voice overdubbing part is always going to be iffy.

    4. I have a lot to say about Descript after a month. I've edited five episodes so far, spent about fifteen hours with their support team, trained two overdub voices, and watched about twenty hours of tutorials, and picked apart help.descript.com.Descript doesn't learn. So you'll constantly be fixing the same thing regarding edit boundaries, word boundaries, gaps, etc. The transcription glossary is helpful but not perfect.Overdub doesn't learn automatically, but a few steps means you can use your regular podcast content to train it with every track you make. You get what you put into overdub.Their help articles don't actually help with anything beyond step 1 of any process. They have no workflows anywhere and you can't see anyone editing anything live. All the YouTube content I can find is paid referrals.Their YouTube content has no timestamps, and everything covers the wow factor of removing text to edit. Their hour long livestreams are very bland, filled with terrible repeated filler words, and the social media manager does way too much talking while the helpful technician takes a backseat in every one I've watchedTranscriptions are done together in compositions, not on individual tracks, so there can be absolutely no cross talk.In the case of cross talk, or descript simply thinking it knows better than you, it will prevent you making changes. Like full stop it will change what you change, back to what's incorrect. Support knows about this.The automated tools don't work great, and you'll need to make manual passes anyway, so you might as well not use them.The stock overdub voices are usable right now in content you're selling, but they've said they might change this in the future, so using them may risk your content.Gap remover removes anything that isn't in a "word boundary" so laughing is gone.The filler word tools also rely on "word boundaries" which aren't frequently accurate, and require the most editing aside from dead air, I've found. Again, descript doesn't learn as it goes, so these changes will be constant and repeated.Overdub can't handle accents. My British co-host has done the 90 minute script and I've ran a few episodes in to train it, and he's come out very comically American. It's actually changed his voice. We all find this very amusing.If you're running a single person, narrative podcast, I can see huge, huge benefit to Descript. But if you're having a conversation, or more than one person in a room- all their YouTube content is using zoom- you'll have big issues. As in, 90% of descript isn't usable.Transcripts are done on the source track, so Studio sound and mic bleed remover don't help with accuracy.Their discord isnt very helpful- usually advising to file a support ticket. There's a message every few days usually.Livechat support hours are 9-5ish PST. Email responses can take a week or more.There are no best practices listed anywhere 18 they don't have device compatibility listed anywhere. The podtrak p4 can't record multitrack into descript for example.They have some odd ui choices regarding the timeline editor, ctrl+alt+e, and starting and stopping playbackDescript transcribes each track separately, so if your podcast is 30 minutes with three people, that's 90 minutes of transcription time you're being chargedIf you want to see what the program is missing, check out their feature request page. It's got 1500 articles of requests by users.Ultimately, I haven't got a single usable transcript, so I'm paying to slide words around and use studio sound. Which, thankfully is very handy compared to waveforms and making these changes manually in a program like audacity. Do I think it's worth $30 a month for that, forever, though? Great question.
  2. Jun 2023