The following post is created entirely by AI. I took the comments from Day 5 of the 12DoAI on translating videos and asked Claude to summarise them. I then used Recast to turn the summary into a podcast:

https://app.letsrecast.ai/r/de1177fb-350a-4d53-888c-371071b310e0

The podcast contains comments and reflections from various people on their experience using AI video translation tools, specifically HeyGen. Most tried recording a short video in English then getting the tool to translate it into another language.

The first person translated their video into Finnish for a Finnish colleague. The translation was very good, just missing a few words here and there. The second person had their video translated from English into Greek. They found the translation quite accurate, though some grammatical errors, and their email/pronouns weren’t translated properly. The context was also slightly altered.

The third person found the process easy but didn’t get to the translation stage due to lack of credits. They could see the value for lecturers with international students. The fourth person echoes privacy and ethical concerns about AI companies’ use of personal data.

Others had technical difficulties registering with HeyGen. One used Rask instead but could only translate 1 minute videos in the free version. The translation seemed good at adjusting speakers’ voices, but they wondered how it would handle complex rapid interactions. Another had fun applying a French voice to English text rather than translating it.

Someone else noticed changes when they used ElevenLabs to create a text-to-speech track. One person used up their HeyGen credits experimenting with voices and languages before creating a video. Another uploaded their own video and was impressed HeyGen translated it into Dutch in their own voice.

One person found HeyGen less easy to navigate than HourOne. The translation was useful but imperfect. They created an intro video but found their animated avatar odd. One person’s Greek colleague confirmed a HeyGen translation from English into Greek was frighteningly accurate and spot on.

The final person points out the French voice over example does not match the presenter’s voice. Overall, people found promise and some impressive capabilities in the AI translation tools but also limitations around context, complex interactions, and ethical concerns over data usage.