Just one tip, since you were going for realism in the second example: make sure the woman's microphone isn't levitating. I'd also dial up the strength of her mouth movements, and have them smiling, since the voices sound like they'd be smiling.
Amazing Video I love how you didn't give away the punchline at start but still started with some powerful examples. Ended up being a pleasant surprise to see you take it to next level. Great thinking.
How did you automatically split the voices into two tracks? I only see the colour highlighting in Descript, but no way to export each speaker track separately.
So it's combining the script and vocal emotion of Notebook LLM, the facial expressions of Hedra, the realistic portraits of Vidu Studio, and the voices of Eleven Labs. It's very cool, but it definitely makes one long for a future where it's all provided by the same app.
@@jason_v12345 No doubt it will happen. Or at least close. But even today in traditional filmmaking, multiple tools are often used to create an effect, so it’s not like anybody has “failed“ if they can’t get one tool to do everything. Thanks so much for watching!
Great video! And thanks for your work! Very similar to the workflow we have been using for our Real talk with AI channel (we went the Hedra route because we had watched your prior videos on that)! We just started up our channel this week. :) We talk about the latest in the AI UA-cam space and will feature your video from today, soon. We have featured others that you have done as well. We love what you are doing. :)
@@BobDoyleMedia You are welcome! The episode is now live (Ep. #5). Check out what AI had to say about your episode (I bet you get a kick out of it). :) Your content has been featured in episodes #2, #4, and #5 now. Our AI podcasters find your content...fascinating. It is almost if they are connected, in a very tangible way, to the very stuff, you have been talking about. ;)
@@RealtalkwithAI-AI So cool! And fast turnaround! I mean I JUST posted the video they are talking about! ua-cam.com/video/95pea0605GM/v-deo.htmlsi=kzv6mcvtOqmvuuUr
@@BobDoyleMedia Thanks! :) I've been trying to optimize my process so that I can respond as quickly as possible (still a good bit more to do in that regard). My general goal with each episode, is that most of them be based on something popular on UA-cam in the AI space, where each episode has a video (or videos) as source inspiration. Ideally, the source inspiration video(s), at my "go public" time, would have been posted from within 24 hours of when my episode goes public. I also plan to include the source videos that inspired each episode in the description. That helps me contribute back to the source inspiration, it helps enrich my episode by pointing to additional rich information and content, and then hopefully it will also be such, as to help build community around our shared interest. :) I'll see how long I can keep that up (given that I also have a family and a separate fulltime job). Optimizing everything helps. :)
@@BobDoyleMedia just change the audio in Capcut pro. But splitting them out is a pain in the toosh. Wish you could export them as multi track. Maybe there is an ai for that...like Stemz or Moises.
Wow. Oh man, that was incredible. You made it look so easy.I just subscribed and will have to rewatch a few times, to make sure I get ir. Thanks so much!!
Basically the last few videos were great bob doyle advertisements. I tend to forget names of content creators, but with songs about bob doyle, podcasts about Bob Doyle I will now remember it haha
I'll watch this tomorrow, but I just wanted to right now, thank you so much for the helpful tutorial. I really appreciate your works and helpfulness to our community. 💪
The problem is it is very tedious and time-consuming to separate the audio, you would have much better luck doing this using 11 labs isolator, or having GPT 01 write you software that can separate the two different frequencies since we know they will almost always be in the same range given the male and female voices. I'm actually working on something similar to this myself. Heygen makes for a much less convoluted workflow btw.
I don't think the 11 labs isolator works in separating different voices. Just the voices from background noise. Any other idea or software that could do this automatically?
Hey Bob, great informative video, as usual. I was wondering how did you create the split screen affect, with the sound waves for the Hedra Podcast clip? Can I do this in CapCut or Canva?
I appreciate the effort you put into this, but I think there might be room for improvement............... It takes quite a bit of time when the audio is around 10 minutes, and talking photos feel a bit outdated with where AI is today. It would be great if the podcast had a more realistic feel. That being said,,,,,,,,,,,,,,, thanks for sharing the video!
I just stumbled onto this channel. Looks sick... I'm very excited to dig in. Is there a video about making the 3d animated Bob Doyle avatar you use? Its very impressive
Hey your video is awesome !!! thank you so much ! but I do have a question, do you think there is a chance to AI tool that can separate the audio file automatically ? Im ready to pay for it
great video! still manual splitting of the two voices is not practical...i am trying to find an automatic way but still no luck. used gemini to create a time stamped transcript and it is doing GREAT job in flagging who is actually talking each time / man or woman! but then i am stuck...any ideas?
Should be possible to create a python script to use your timestamps and cut the video and merge all the parts with male only and female only. Could use Claude to help create to script.
I thought of doing it that way but is there a way to bump up the accuracy or body language and mouth movement, so it is closer to Hedra? Also with a little work you should be able to composite both together as if they were on the same set. I only did it with a single person where the background was extended to a full video dimensions. just hope Hedra gets a update. It really needs some higher resolution output and the ability to do it with different angles. If you didn't want to go through the complete changing of the voices you can just do some pitch shifting. New voices would be better but if you don't want to put the time into it that is an option. I also want to find if there are ways to modify the notebooklm podcast to trigger certain things. Say like make a intro and outro for the podcast with a name of the show etc.
Could you clarify whether the Audio Output Podcast feature can be used in a commercial setting? Specifically, if I input data that I own or have permission to use, and NotebookLM generates an audio podcast, would I be able to post that content on my website or other platforms for commercial use?
Thanks for sharing. It must have been a lot of work. I would just change the lipsync, which is a bit static. I tried changing the language of a podcast wav file, but it didn't even work out. And dubbing services are pretty expensive.
nice im actually working on a flow that uses a comfy TTS. so can clone my or other peopels voices in. I think we could also use audimees voice bank..(since its paid for) But the biggest pain is the time to render live portrait. Oh and I would also use a web video of me dubbing lipsyncing or mouthign the script for the liveporttrait source movment video.
@@victoriadedicova No, it's definitely not Free and Unlimited. $10 for 20 min / mo, $50 for 110 min / mo and more restrictions like the lenght of the video ... etc.
Do you know if there is anything to do this with 2d images like say cartoon faces or characters? If you could animate a drawing from audio that would be cool.
How do you get the two avatars to speak in order? The video Hedra creates is straight through based on the cuts made on audacity. How do I make it as if they were speaking to each other from the two different audio file and two different avatar videos?
Awesome Bob. I love your channel. Do you have any link to live portraits on MimicPC? I know you have a video about that service. Make it an affiliate one if you can so I can support you 😊
Is there an automated way to do the workflow that was done in audacity? Trying to look for a software that can do the separation of voices in 2 audio files preserving the time the speakers speak. thanks! :)
I have descriptive, and I’ve been trying to figure out how to do that because I keep hearing that you can. It obviously will separate the speakers, so it knows how to do certain things, but I have not figured out how to export each audio track individually
Did you use Hedra to generate your Avatar which you used to talk in the video? Can you suggest us a free tool or a tutorial to become a Vtuber with free platform
@@BobDoyleMedia If you find anything please let us know. I hoped AudioStrip could handle it but they only seem to be able to split lead and backing vocals
Great video, but how did you create the avatar of you animated and all. I am new to your channel. I would love to create an avatar for myself to make vids
This is pretty interesting, but, I found some halucination to be an issue. Date changes and such. For example it changed a birthdate of November 1917 to born in 1918. Know your material.
Thanks, Bob, I am a professor and appreciate what you put on UA-cam, to create better and more engaging material for students. I am in Brazil, and as NotebookLM only produces audio in English, I had a lot of work to make it Portuguese. I downloaded the audio and generated the script using TurboScribe. Then I used chatGPT to translate it into Portuguese. Then I run the script by a text-to-voice generator twice, one with a female voice, and the other with a male voice. After that, In audacity, I clipped from each track the parts that were for each speaker, leaving the male track with only his parts, and the female track with her parts. It actually works, but the "conversation" wasn´t as natural as the one NotebookLM generates, which is a bummer, since that is a great interaction. Do you have other methods to do it? is there a service that automatically translates the audio, without the need to regen it from a script?
Why would you do something like that foe your students?!it takes too much time and for now that s just a gimmick that adds no value. You should make better use of your time than that. Explain your students what embeddings/vectorization , do it well and you would create more value than useless podcasts.
I recommend ElevenLabs' dubbing feature. Essentially, it takes the audio, clones the two voices, and generates similar voice clips to the original English ones. It then sends you the audio file back, but translated into Portuguese by the same NotebookLM voices and with all the natural quality of the conversation. This method takes less than 5 minutes. Hope this helps!
You can actually automate the face animation in comfyui with a workflow that takes audio and animates a talking head with Anitalker, then convert that to the actual head with liveportrait all in a single process.
Actually, seperating into 2 vocals (male and female) is not an easy task on Audacity.. There must be AI tools to do that automatically, because AI Dub services like ElevenLabs and Speecify detects the number of speakers and add different dub voices for each.. so, there must be a easy way..
Dude..I have been trying to do this for the last 2 days and you just helped me get there so much quicker...thank you thank you thank you!!😄
@@UnclePapi_2024 so glad to hear it!
i wanted to ask u brother, where are you uploading ur podcasts? since youtube wont monetize ai podcast.. thank u in advance !
Just one tip, since you were going for realism in the second example: make sure the woman's microphone isn't levitating. I'd also dial up the strength of her mouth movements, and have them smiling, since the voices sound like they'd be smiling.
Holly molly, that's a day of work just to split voices dude!, we are in the AI age where you just with one prompt to split voices!
Amazing Video I love how you didn't give away the punchline at start but still started with some powerful examples. Ended up being a pleasant surprise to see you take it to next level. Great thinking.
@@timba2647 Thanks so much!
I did something similar by uploading the audio file to Descript and used their auto speaker detect to automatically split the voices into two tracks.
How,? I tried doing this but couldn't get it to do that.
@@user-fp1lr2ip1x I also tried that and couldn’t figure it out.
Interested to know how you managed to do this as well
I as well.
How did you automatically split the voices into two tracks? I only see the colour highlighting in Descript, but no way to export each speaker track separately.
So it's combining the script and vocal emotion of Notebook LLM, the facial expressions of Hedra, the realistic portraits of Vidu Studio, and the voices of Eleven Labs. It's very cool, but it definitely makes one long for a future where it's all provided by the same app.
@@jason_v12345 No doubt it will happen. Or at least close. But even today in traditional filmmaking, multiple tools are often used to create an effect, so it’s not like anybody has “failed“ if they can’t get one tool to do everything. Thanks so much for watching!
Your unique perspective always adds value!
Great video! And thanks for your work! Very similar to the workflow we have been using for our Real talk with AI channel (we went the Hedra route because we had watched your prior videos on that)! We just started up our channel this week. :) We talk about the latest in the AI UA-cam space and will feature your video from today, soon. We have featured others that you have done as well. We love what you are doing. :)
Wow, thanks so much! I really appreciate all that!
@@BobDoyleMedia You are welcome! The episode is now live (Ep. #5). Check out what AI had to say about your episode (I bet you get a kick out of it). :) Your content has been featured in episodes #2, #4, and #5 now. Our AI podcasters find your content...fascinating. It is almost if they are connected, in a very tangible way, to the very stuff, you have been talking about. ;)
@@RealtalkwithAI-AI So cool! And fast turnaround! I mean I JUST posted the video they are talking about! ua-cam.com/video/95pea0605GM/v-deo.htmlsi=kzv6mcvtOqmvuuUr
@@BobDoyleMedia Thanks! :) I've been trying to optimize my process so that I can respond as quickly as possible (still a good bit more to do in that regard).
My general goal with each episode, is that most of them be based on something popular on UA-cam in the AI space, where each episode has a video (or videos) as source inspiration. Ideally, the source inspiration video(s), at my "go public" time, would have been posted from within 24 hours of when my episode goes public. I also plan to include the source videos that inspired each episode in the description. That helps me contribute back to the source inspiration, it helps enrich my episode by pointing to additional rich information and content, and then hopefully it will also be such, as to help build community around our shared interest. :)
I'll see how long I can keep that up (given that I also have a family and a separate fulltime job). Optimizing everything helps. :)
@@BobDoyleMedia just change the audio in Capcut pro. But splitting them out is a pain in the toosh. Wish you could export them as multi track. Maybe there is an ai for that...like Stemz or Moises.
Wow. Oh man, that was incredible. You made it look so easy.I just subscribed and will have to rewatch a few times, to make sure I get ir. Thanks so much!!
NotebookLM and ElevenLabs are great. The Hedra tech still has a long way to go in terms of lip sync and facial expressions.
Awesome but I still love the hedra clip
THIS is exactly what I was thinking for my YOutTube channel I have been planning ! WOW thanks for this video! ;D
❤you just put it all together 😊
Like I worked on for 2 years with hardly any technology available.
Thanks
I can't believe him.. Everything I need in one download
Basically the last few videos were great bob doyle advertisements. I tend to forget names of content creators, but with songs about bob doyle, podcasts about Bob Doyle I will now remember it haha
I'll watch this tomorrow, but I just wanted to right now, thank you so much for the helpful tutorial. I really appreciate your works and helpfulness to our community. 💪
The problem is it is very tedious and time-consuming to separate the audio, you would have much better luck doing this using 11 labs isolator, or having GPT 01 write you software that can separate the two different frequencies since we know they will almost always be in the same range given the male and female voices. I'm actually working on something similar to this myself.
Heygen makes for a much less convoluted workflow btw.
I don't think the 11 labs isolator works in separating different voices. Just the voices from background noise. Any other idea or software that could do this automatically?
Until I can pay someone overseas to get the work done I'm just going to stick to the podcasts.
The young lady has a floating SM7B - I want one like that!😆 But thanks Bob, you're my tutorial source for Ai tools!
If you ad Wave2lip to the Workflow after the video has been generated it will help
Is that hair, or a secret medieval weapon? 🤔 Awesome video! It's amazing on how believable the AI podcast comes across. Dang.
Hey Bob, great informative video, as usual. I was wondering how did you create the split screen affect, with the sound waves for the Hedra Podcast clip? Can I do this in CapCut or Canva?
I appreciate the effort you put into this, but I think there might be room for improvement............... It takes quite a bit of time when the audio is around 10 minutes, and talking photos feel a bit outdated with where AI is today. It would be great if the podcast had a more realistic feel. That being said,,,,,,,,,,,,,,, thanks for sharing the video!
Tons of fun as always, Bob :D
Excellent points, professor!
I just stumbled onto this channel. Looks sick... I'm very excited to dig in. Is there a video about making the 3d animated Bob Doyle avatar you use? Its very impressive
Or is Hedra used for that as well
I actually followed that and will have a go. Thanks for making this not seem so overwhelming!
NotebookLM is so great, and unique tool with great value.
Love the close threat! LOL!
Loving the AI hair. Def need that in RL.
Simply brilliant! Thank you for sharing!
Hey your video is awesome !!! thank you so much ! but I do have a question, do you think there is a chance to AI tool that can separate the audio file automatically ? Im ready to pay for it
great content bob
great video! still manual splitting of the two voices is not practical...i am trying to find an automatic way but still no luck. used gemini to create a time stamped transcript and it is doing GREAT job in flagging who is actually talking each time / man or woman! but then i am stuck...any ideas?
I'm trying to figure out the same... Chime back if you figure it out and I'll do the same!
Should be possible to create a python script to use your timestamps and cut the video and merge all the parts with male only and female only. Could use Claude to help create to script.
I thought of doing it that way but is there a way to bump up the accuracy or body language and mouth movement, so it is closer to Hedra? Also with a little work you should be able to composite both together as if they were on the same set. I only did it with a single person where the background was extended to a full video dimensions. just hope Hedra gets a update. It really needs some higher resolution output and the ability to do it with different angles. If you didn't want to go through the complete changing of the voices you can just do some pitch shifting. New voices would be better but if you don't want to put the time into it that is an option.
I also want to find if there are ways to modify the notebooklm podcast to trigger certain things. Say like make a intro and outro for the podcast with a name of the show etc.
How do you do speech to speech for larger audios (11-15 mins)? the voice changer feature doesn't allow files above 5 minutes. Thanks!
You just have to do some editing in your video editor. Fairly routine. Just create multiple files and put them together in another program.
Very cool. It’s a lot of work tho’. I hope Google Notebook LM will include all of this before very long 😆
Thank you for makimng it simple,
Bob this is amazing what an amazing video....thank you
Could you clarify whether the Audio Output Podcast feature can be used in a commercial setting? Specifically, if I input data that I own or have permission to use, and NotebookLM generates an audio podcast, would I be able to post that content on my website or other platforms for commercial use?
Thanks for sharing. It must have been a lot of work. I would just change the lipsync, which is a bit static. I tried changing the language of a podcast wav file, but it didn't even work out. And dubbing services are pretty expensive.
nice im actually working on a flow that uses a comfy TTS. so can clone my or other peopels voices in. I think we could also use audimees voice bank..(since its paid for) But the biggest pain is the time to render live portrait. Oh and I would also use a web video of me dubbing lipsyncing or mouthign the script for the liveporttrait source movment video.
I love the video but am I missing something, Hedra only gives you videos of 1 min long for $10 a month, $50 a month for 4 min long videos, ouch!
It's free and unlimited
@@victoriadedicova No, it's definitely not Free and Unlimited. $10 for 20 min / mo, $50 for 110 min / mo and more restrictions like the lenght of the video ... etc.
@@victoriadedicova
Genuinely curious, what do you gain from spreading lies?
?! lies?
@@NakedSageAstrologyit’s a bot
How do we change the voices to our own voices instead of what just given?
We do that at the very end of the video. :)
Could you just throw that into descript and it will separate out the two speakers for you yea?
What tool did you use for your ‘forgot to turn my camera on’ avatar? Hedra?
Yes. :)
What video editor did you use for the split screen effect?
Do you know if there is anything to do this with 2d images like say cartoon faces or characters? If you could animate a drawing from audio that would be cool.
How do you get the two avatars to speak in order? The video Hedra creates is straight through based on the cuts made on audacity. How do I make it as if they were speaking to each other from the two different audio file and two different avatar videos?
What software or tool was used to get that avatar with the spikes for an audio to video? Was it Hedra too?
How do i add the avatar to explainer videos
Men the whole video, I loved your character mostly.....😂
Complicated I wish one click solution exists anyway thank for creativity
What software application did you use for the animated version of yourself?
Awesome Bob. I love your channel. Do you have any link to live portraits on MimicPC? I know you have a video about that service. Make it an affiliate one if you can so I can support you 😊
Is there an automated way to do the workflow that was done in audacity? Trying to look for a software that can do the separation of voices in 2 audio files preserving the time the speakers speak. thanks! :)
Hindenburg
Great! Do you know when are the interviews can be created in spanish?
I think you may be able to split the voices in descript
I have descriptive, and I’ve been trying to figure out how to do that because I keep hearing that you can. It obviously will separate the speakers, so it knows how to do certain things, but I have not figured out how to export each audio track individually
What’s the avatar AI for using your own face for an avatar ??
Is there an AI tool for Audacity that does the male / female separation???
Did you use Hedra to generate your Avatar which you used to talk in the video? Can you suggest us a free tool or a tutorial to become a Vtuber with free platform
I wonder if my audio software could split the voices automatically like it does for music creating stems for each.
@@mjfII I spent some time looking for a quick ai solution for this, as I’m sure it exists, it I just ran out of time and did it old school. 😎
@@BobDoyleMedia I asked Gigi (chatgpt)..she mentioned 'Speaker Diarization' but I haven't moved that far ..yet 😏
Descript can detect voices ✅
@@BobDoyleMedia If you find anything please let us know. I hoped AudioStrip could handle it but they only seem to be able to split lead and backing vocals
love this! could you do a video on HEDRA?
I've done several, including just this week. Here's the latest! Thanks for watching! ua-cam.com/video/mkz6eAK-uak/v-deo.htmlsi=W4pi7whxg_6OXhhj
Great video, but how did you create the avatar of you animated and all. I am new to your channel. I would love to create an avatar for myself to make vids
Actually, the video itself describes exactly how I did that. I used Hedra, which is what I used to animate the Podcasters in the first example.
Nice to be able to do this but I'm guaranteeing this time next year video and face replacement will be available. Voice also for a fee.
This is pretty interesting, but, I found some halucination to be an issue. Date changes and such. For example it changed a birthdate of November 1917 to born in 1918. Know your material.
Awesome. Is it possible to translate it to another language?
Is it possible to monetize that? what do you think?
For now
Descript might help out
This really helped
Great to hear! Thanks for watching!
Great video
Great work
at the last test, did you manually separate the man's voice from the woman's?
@@tokyofamily8536 yes, all tests were done with the separated voices I edited in the video.
Good job done
@@LofiBeatsMusic4u thanks so much for watching!
I subscribed 😂
Can I hire your to set this up for me?
did you try Kling ai?
Is there a a way to do this automatic "Audio diarization" ? 04:00
Hedra only support 40 to 45 sec...!
It supports almost 5 min of generation
Great video - I'm interested in how you created your animated avatar for the bit you didn't record. Have you covered that in another video somewhere?
I cover it in THIS video. I used Hedra, the first platform I demonstrated. :)
@@BobDoyleMedia cool - doh! - I think I was distracted by the avatar's look. I'm assuming that's AI generated. it's looks very good.
@@BobDoyleMedia Which Stylized feature ?
@BobDoyleMedia
Can you please make a video on how to make that avatar and add audio to it and use it in explainer videos. Thanks
Good work!!
the only thing lacking is changing the two persons audio!
I cover that specifically on this video: ua-cam.com/video/sk3MtYj0tMI/v-deo.htmlsi=M-UPBim2SQ3RAqi9&t=644
That s a nice avatar, how did he do it?
I hope podcast feature is available in spanish soon.
I ❤ your fake hairstyle
@@autonicaadabsurdum Thanks! 🥸
A little tedious but very helpful.
Super !
Fantastic!!!
Just use Canva to split audio automatically
is this the same how it was done in audacity but this time it is automatic in canva?
Thanks, Bob, I am a professor and appreciate what you put on UA-cam, to create better and more engaging material for students. I am in Brazil, and as NotebookLM only produces audio in English, I had a lot of work to make it Portuguese. I downloaded the audio and generated the script using TurboScribe. Then I used chatGPT to translate it into Portuguese. Then I run the script by a text-to-voice generator twice, one with a female voice, and the other with a male voice. After that, In audacity, I clipped from each track the parts that were for each speaker, leaving the male track with only his parts, and the female track with her parts. It actually works, but the "conversation" wasn´t as natural as the one NotebookLM generates, which is a bummer, since that is a great interaction. Do you have other methods to do it? is there a service that automatically translates the audio, without the need to regen it from a script?
Why would you do something like that foe your students?!it takes too much time and for now that s just a gimmick that adds no value. You should make better use of your time than that. Explain your students what embeddings/vectorization , do it well and you would create more value than useless podcasts.
Have you tried HeyGen?
I recommend ElevenLabs' dubbing feature. Essentially, it takes the audio, clones the two voices, and generates similar voice clips to the original English ones. It then sends you the audio file back, but translated into Portuguese by the same NotebookLM voices and with all the natural quality of the conversation. This method takes less than 5 minutes. Hope this helps!
@@TBS6217thank you! I knew there was dubbing ai but didn’t know where
Very great
nice video
Amazing
Is it free or paid?
The bots are way too sugar sweet. Almost too much to seem legit. Wish one could dial the chumminess in/out
How can you make it just do a single voice talking about the subject? I dont want the female at all
probably get the transcript using gemini and then use text to voice
You can actually automate the face animation in comfyui with a workflow that takes audio and animates a talking head with Anitalker, then convert that to the actual head with liveportrait all in a single process.
All they had to do was produce a stereo track with the voices in their own channel. Guess they never thought of it.
They have a feature request on their discord channel
@@yoagcur Sweet. I'm way too lazy to do what Bob makes look easy here.
@@bobhawkey3783 But are you too lazy to make the feature request?
Actually, seperating into 2 vocals (male and female) is not an easy task on Audacity..
There must be AI tools to do that automatically, because AI Dub services like ElevenLabs and Speecify detects the number of speakers and add different dub voices for each.. so, there must be a easy way..
@@ZenAIMusic I agree.
Phenomenal
yeah but on a 22-40 minute podcast it will take forever
@@Justin-General I’m sure there’s a solution out there.
good