So… I can upload a video by providing a link but Pickaxe tells me it cannot analyse the content. I can get this analysed very effectively by ChatGPT (all models).
So… I implement an action to chain this and send it to OpenAi and specifically get it analysed… this seems to start and then I get the routine message that there isn’t any functionality to access videos… Why? Its a link, why can’t I pass this to OpenAi?
Has anyone figured out a workaround?
Thanks
It’s just a matter of prompting it. It can definitely summarize videos. What’s the prompt you’re using?
‘Analyse the content of this video of an athlete and give me feedback on issues, problems and improvement opportunities for their form. Also provide a detailed plan to remedy issues to include strength and conditioning: (video link)’
I’ve tried videos via youtube, own open server and via dropbox… The response is always the same: “It seems I cannot access or analyse content from external links or videos directly. However, you can provide me with specific details about the content of the video, including aspects of the athlete’s performance, any observed movement patterns, or particular concerns you have.”
Of course, pasting the same response into pretty much any GPT gives me exactly what I am looking for but I cannot get this to work.
I had the same issue… the problem is most likely API related… For whatever reason, OpenAI limits capabilities depending on your API Tier! I couldn’t find official information on this, but I am still on Tier 2 (just started) and if I put my API key I have the same problem… If I select “Use Pickaxe Credit” it works… the difference is just the API-Key and my assumption that Pickaxe has made it to the highest tier! 
Thanks for the suggestion @easydeutsch - unfortunately, this makes no difference (although it makes failure a little quicker)!
I receive largely the same message when using pickaxe credits:
“I am unable to access the video through the provided link, as it seems to lead to an unsupported format or contains corrupted data. To assist you in analysing the athlete’s performance effectively, I would need a functioning video link that is compatible with standard video formats.”
The video is an mp4 fully tested and the link is the same unprotected link used directly with OpenAi to check the comprehensive response received when used directly in ChatGPT so I know it its not corrupt or unsupported.
I would recommend that you check out the “Content Summarizer” template from the examples in the builder. Then build your prompt up on top of that solid foundation.
Not sure that I’d describe this as a solid foundation as it still suffers the fundamental problem as evidenced by the response:
“I’m unable to access external links or analyze videos directly. However, if you can describe the content of the video or mention specific aspects of the athlete’s performance you’d like feedback on, I can help you analyze it and provide suggestions for improvement.”
As with all attempts, it claims it can’t access the video… it doesn’t appear to be passing the link to the video to the underlying AI model or at least, doesn’t think it is capable of doing so.
I have also noticed that one of my very reliable pickaxes has now stopped working claiming that it can’t access public strava profiles. As with the video issue, when I pass the strava profile directly to a GPT, it works just fine and can analyse an athlete profile. This was working until last week.
Huh, that’s quite odd. It does not do that for me. Are you giving it a youtube link? Are you adding more prompts? Did you change the model?
With the template, you should be able to drop in any youtube link and it will process it. I haven’t found one it doesn’t work for. Be careful about adding tons of instructions talking about accessing videos. You will confuse the model because language models don’t think they can access videos. Video-scraping is an ability we built on top of it.
I provided an mp4 stored on an open server with a direct link. I have one line about accessing the video then have a prompt and rules to shape the response. I tried the template you signposted and left it only with the most basic prompt modified for video and got the response I posted above.
also… occasionally getting this error:
An error occurred with your deployment
EDGE_FUNCTION_INVOCATION_TIMEOUT
lhr1::chs62-1739925246746-c6acb5cbe05e
I’m having a similar issue with a video analyzer. It seems that the AI will pickup the transcript and audio. But it can’t “see” the video. Specifically it has an NA for “Visual Aids & Slide Quality”
My results are:
Overall Presentation Rating: 8/10
Detailed Feedback:
Content Clarity & Structure: 9/10
Feedback: The content of the presentation is well-organized, with a clear structure that guides the audience from the problem (chaos in business) to the solution (decision-making frameworks). The presenter effectively introduces personal anecdotes to illustrate points, making the content relatable. However, emphasizing key messages with clear takeaways or a summary at the end could reinforce the presentation’s effectiveness.
Delivery & Engagement: 7/10
Feedback: The speaker uses personal stories, like the yacht incident, to engage the audience, which is effective in capturing attention. However, it would be beneficial to know if the delivery includes varying vocal tones and body language to maintain engagement throughout. If possible, incorporating more dynamic delivery methods or interactive elements could enhance audience engagement.
Pacing & Timing: 8/10
Feedback: The pacing appears to be well-balanced, with the presenter covering each point thoroughly without rushing. To further enhance pacing, consider incorporating pauses for emphasis or to allow the audience to reflect on key points, ensuring they have time to absorb the information presented.
Visual Aids & Slide Quality: N/A
Feedback: Since visual aids or slides are not mentioned in the transcript, it’s unclear how they were utilized. If they are part of the presentation, ensure they are clear, visually appealing, and complement the spoken content. If not, incorporating them could add an additional layer of clarity and engagement.
Overall Impact: 8/10
Feedback: The presentation effectively communicates the importance of structured decision-making frameworks in navigating business chaos. The use of credible sources, such as the Stanford decision quality process, adds authority to the content. To maximize impact, the presentation could benefit from a compelling call-to-action or practical examples of how the audience can implement these strategies in their own contexts.
Additional Observations:
