Otis speaks about the Fourth of July.

Stars. Stripes. Sunglasses. A tiny American flag. 

Estimated reading time: 5 minutes

Today also marks another kind of independence.

For the past year, Otis has lived in the pages of The Gospel According to Otis, offering his weekly observations on creativity, work, discipline, and the occasionally puzzling behavior of human beings.

Today, he finally opens his own mouth.

This is the second attempt at presenting Otis Speaks.  We haven’t got all our ducks in a row, but who does? Quack.

Like every first creative project, it comes with fingerprints still drying. I am going to turn this over to Jeff now. In his own words and the process. 

He has the brains. And I am out of my pay grade here. This is just in case anyone out there wants some insight into the process. 

Jeff Fried Speaks

Starting point:

GPT-generated image and the script was supplied by Bryan I used a previously created Otis character created on OpenArt, Otis voice created on Eleven Labs.

Video Generation:

First, I generated the Otis voice speaking the script.  This gave me 23 seconds of audio.  Since video models can only generate up to 15 seconds, I split the audio into 11- and 12-second segments. I then prepared a reference image for the first segment by asking OpenArt to replace the character in the image Bryan gave me with my OpenArt Otis character. 

I chose the Seedance 2.0 mini model to generate the video.  This yields results similar to the full Seedance model but costs half as many credits.  I then generate two clips: Otis_4thOfJuly_1 and Otis_4thOfJuly_2.  

For Otis_4thOfJuly_1, I used a reference image from Bryan with my Otis character replacement and the 11 second audio segment created in Eleven labs. Here is the prompt used, where @Image1 and @Audio1 are the reference image and audio referred to in the previous sentence.

Prompt:

The cat in @Image is talking to the camera saying, using the voice and inflection in @Audio1 “Here you go—funny stuff—Every year, humans celebrate freedom by declaring war on the atmosphere. Loud explosions.
Flashing lights.
Smoke everywhere.”  In the background are exploding fireworks and a sparkler in the foreground.  The sound of the fireworks and patriotic marching band music are heard at a low volume in the background

For Otis_4thOfJuly_2 I used the ending frame from Otis_4thOfJuly_2 and the second 12 second Audio segment from Eleven labs referred to as @Image1 and @Audio1 in the following prompt:

The cat in @Image1 is talking to the camera saying, using the voice and inflection in @Audio1 “Then they wonder why every dog, cat, and squirrel in the neighborhood is hiding under the furniture.

You call it a holiday.

We call it an invasion.

Wake me when you’ve learned to celebrate with tuna.”  The sound of the fireworks and patriotic marching band music are heard at a low volume in the background

Editing:

I used DaVinci Resolve to join the clips into a single 23 second video.  No transitions or effects were used.

Problems: There two main problems in the video. The first is that much of the British accent and cynacism of the Eleven labs created Otis voice is missing in the actual voice generated from video.  This was the case both using the Eleven labs generated audio as well as the voice assigned to OpenArt character Otis which was generated using the Eleven labs Otis voice.  This is partially rememadied by amending the prompt to exaggerate the accent and cynacism of the voice in the prompt. I also attempted to use OpenArt’s lip-sync tool, but the only model that works on video does not recognize non-human characters.  I did not attempt the image only models as I’m unsure how they’d handle the rest of the animation and I didn’t want to burn credits experimenting.  This is a topic for further research.

The second problem is a slight discontinuity between the first and second clips.  It is barely noticeable, but it is a known problem even in the full Seedance 2.0 model.  My usual solution is to change the camera angle or, if possible, transition to a different scene, avoiding any continuous fixed camera angle for more than 15 seconds. I have not attempted that here because the discontinuity is barely noticeable and we’re still experimenting,

---

Can We Help You?

Thank you for following Otis from the written page to the animated screen. If you enjoyed this adventure, please like, comment, share, and subscribe to Otis Speaks. We’d love to hear what you think as Otis’s world continues to grow.

Want More Mack-n-Cheeze?

Videos - Bryan At Mackncheeze on YouTube

Podcasts – Bryan At Mackncheeze Apple Podcasts, Fountain, Spotify