AI Generated Anchors with FAST Channels based on VOD2Live
With this blog post we want to demonstrate the idea of using the latest advancements in open AI technology to add AI generated anchors in automatically generated FAST channels.
Not only does the VOD2Live technology [1] avoid redundant transcoding and packaging but it also enables you to scale up with numerous FAST channels at a low cost. Based on existing VOD content library automatically generate schedules to offer a big plate of FAST channels and increase the return of content investment. With the AI technology available today we can enhance this experience by adding an AI generated anchor between the programs.
To demonstrate this idea, we created a VOD2Live channel based of Big Buck Bunny, Tears of Steel, and a presentation from Streaming Tech Sweden 2022. The manuscript for the anchor to read was generated simply by asking ChatGPT [2] to generate an introduction based on title and synopsis where available. For Big Buck Bunny and Tears of Steel no synopsis was needed.
The anchor videos were generated using the online tool AI Studios [3], a Text to Video Generation Platform by Deepbrain, based on the manuscripts from ChatGPT and a graphics template created in PowerPoint. Below are examples of the AI-generated anchor, one where he introduces Big Buck Bunny and one for the Streaming Tech Sweden talk. These anchor videos were then added to the VOD2Live playlist.
In this case this process was manual to quickly be able demonstrate this concept. The idea is to automate this process based on an automatically generated schedule. For example, we could extend the auto schedulers in our open-source Schedule Service to start this process when a new event is populated to the schedule. Based on metadata and title of an upcoming program it can generate an anchor video that is added to the schedule before the program starts.
The VOD2Live channel we created for this proof-of-concept can be viewed in the player here: https://web.player.eyevinn.technology/?manifest=https%3A%2F%2Ffast.vc.eyevinn.technology%2Fchannels%2Faianchor%2Fmaster.m3u8
Continuation of this proof-of-concept is of course to automate this process and evaluate different AI video generation tools. The intention with this concept is not to replace real-life anchors but rather be able to enhance the experience for FAST channels that would not have any anchors at all.
Eyevinn Technology is world leading independent specialists in video technology, video development and sustainable streaming. Proud organizer of the yearly nordic conference Streaming Tech Sweden.
Further reading on VOD2Live technology
- FAST Channels with Dolby Atmos enabled by VOD2Live technology: https://medium.com/@eyevinntechnology/fast-channels-with-dolby-atmos-enabled-by-vod2live-technology-f11847960b14
- Create your own FAST channels based on VOD2Live technology and Open Source Components: https://medium.com/@eyevinntechnology/create-your-own-fast-channels-based-on-vod2live-technology-and-open-source-components-893197fdea0a
- S06E11: Game of Streams (Too much FAST and content aggregation): https://medium.com/@eyevinntechnology/too-much-fast-and-content-aggregation-41603ab2bac1
References
[1] https://medium.com/@eyevinntechnology/create-your-own-fast-channels-based-on-vod2live-technology-and-open-source-components-893197fdea0a
[2] https://chat.openai.com
[3] https://aistudios.com/
