[Afpif] AfPIF Livestream + Archive.

Wed Aug 27 19:57:47 UTC 2025

Glad to hear it.

I did do some back up recording, but not all. Just in case someone for
got to hit record. But I think Henok was on the ball.

If you do a YouTube Channel I could certainly help you with some legacy
events. The last few are on the Internet archive.
https://archive.org/details/@isoc_live?query=afpif

One can just sit with a hand on the mute button, but it seems to me,
subjectively perhaps, that YT has been less savage with takedowns on
livestreams, and just muting on recordings. I could be wrong.

At the same time as AfPIF I actually did simulcast the CIGF out of Cuba,
and the AV team there actually constructed their own tune for
interstitials, using the Suno app. https://suno.com/

You can hear it here.
https://youtu.be/XVvhrk7_m7Q?list=PL9-4wsDIxlCCQrl7QjVvm29vY7jyL3Oru - it
reminded of the thing Henok always used to play, which I got thoroughly
sick of :)

My own archiving workflow now includes AI transcription, summarization, and
translated captions.

=================

1) Index sessions with speaker info. Make title frames for each session ,
and one end frame for all.

Then, for each session

2) Doing a quick edit on the video, mainly trim, possibly cropping and
split screening with slides, add title and end frames, and the blank skinny
'lower third' at the bottpm.

3) Running that through Whisper Large V3 with a beam of 15 - and convert to
English. Pretty accurate although GIGO. This is remarkably good at
identifying speakers.

4) Reformatting that transcript in Claude (custom prompt) for smooth flow.

5) Inputting the video to  Descript, and then importing that transcript.

Then in Descript

6) checking and fixing that (This is the laborious bit)

8) Enhance audio a bit and remove dead air and some filler words etc..Add a
'speaker name' text field, which auto adds the name of whoever's speaking.

9) Run AI translation, usually ES/ FR but PT is available

10) Export as mp4, wav, docX (EN), and srt (EN/FR)

11) In MS Word, reformat the DocX and add title frame (smaller jpg), and
export to 'EN.TRANSCRIPT' docX and pdf.

12) Run the EN.TRANSCRIPT docX through ChatGPT (speaker by speaker topic by
topic) for a summary.

13) Replace transcript with that and save as EN.SUMMARY.

14) Upload mp4, wav, pdf, title frame to Internet Archive  (converts audio
to mp3, pdf to ePub).+ lides if available.

15) Upload mp4 to YouTube, add description + links to transcript and
summary, + EN / FR subtitles.

when all are done

16) Post index with links to all files and Internet Archive embed on
https://isoc.live = PERMALINK.

===============================

You can see this work  in progress on the CIGF  -
https://archive.org/details/cigf2025

I realize this all is likely well beyond your capacity, but there's always
a chance it could be used as a guidene for any volunteer help you might
have.

What I am seeing is the AI summaries can be highly useful, especially for
the TL;DR crowd. and they can be created with out going through the entire
editing process. Just putting the audio through the Whisper > Claude >
ChatGPT process.

Note 1) Whisper, at that level is way more accurate than the Otter built in
to Zoom. but 2) Whisper outputs barely formatted text, so that's why you
need Claude (ChaptGPT insists on hallucinating when reformatting, Claude
can be constrained) DeepSeek also works, but a) overworked servers b)
Chinese). 3) However ChatGPT does better at summarizing than Claude, Claude
likes to make flowery prose, whereas ChatGPT will do terse analysis.

I have a portable x64 version of Whisper I can share. Both Cluse and
ChatGPT require accounts if one is doing any volume (as does Descript).

Content in Descript can easily be repurposed to Instagram reels etc.

HTH

Joly

On Wed, Aug 27, 2025 at 2:44 PM Kyle Spencer <kyle at stormzero.com> wrote:

> Hi Joly,
>
> Sorry for the delayed reply on this. We were in the heat of AfPIF planning.
>
> Thank you for the guidance. We followed a very similar strategy for the
> content recording. We're hoping to launch a YouTube channel soon so that
> people can easily re-watch the talks.
>
> One thing I think will be important going forward is to ensure that all
> music played at the conference is Creative Commons so that we don't get hit
> with any copyright strikes.
>
> Regards,
> Kyle Spencer
>
> On Wed, Aug 13, 2025 at 3:59 PM Joly MacFie <jolynyc at gmail.com> wrote:
>
>> Thanks for response, Kyle.
>>
>> If you can, get a good back up recording! And remember,;ultimately it's
>> just the audio that matters. Even a phone on the podium recording mp3 can
>> be a lifesaver.
>>
>> If running Zoom  I will suggest is that you do a good cloud recording,
>> according to my recommended settings
>>
>>    -
>>
>>    *Enable “Record active speaker, gallery view, and shared screen
>>    separately”* –
>>
>>
>>    -
>>
>>    *Record a separate audio file for each participant* (*Settings >
>>    Recording > Record a separate audio file for each participant*) –
>>    essential for clean post-editing and noise removal.
>>    -
>>
>>    Make sure “Optimize for 3rd-party video editor” (in *Recording
>>    Settings*) is enabled so Zoom uses a consistent frame rate and audio
>>    sync.
>>
>> I advise downloading all and clearing storage at the end of each day so
>> you don't go over quota.
>>
>>
>> Joly
>>
>> On Mon, Aug 11, 2025 at 3:51 AM Kyle Spencer <kyle at stormzero.com> wrote:
>>
>>> Hi Joly,
>>>
>>> Sorry we didn't reply earlier. This got lost in the jumble of event
>>> planning.
>>>
>>> We have a very tight budget this year (we're actually running a deficit)
>>> due to the scale of the event and a variety of other factors, so we just
>>> don't have the money to do this.
>>>
>>> Hopefully next year the story will be different and we can engage you
>>> for these services.
>>>
>>> Regards,
>>> Kyle Spencer
>>>
>>>
>>> --
>>> Cell / WhatsApp / Signal: +256786350675
>>>
>>> On Mon, 11 Aug 2025, 10:43 Joly MacFie, <jolynyc at gmail.com> wrote:
>>>
>>>> Just following up. Since I have heard nothing, I assume my services are
>>>> not required.
>>>>
>>>> If there is any similar effort to properly publish and archive AfPIF
>>>> 2025 planned, please let me know and I will publicize.
>>>>
>>>> Joly.
>>>>
>>>>
>>>>
>>>> On Thu, May 15, 2025 at 7:08 PM Joly MacFie <jolynyc at gmail.com> wrote:
>>>>
>>>>> Hi Kyle, Gaelle, all.
>>>>>
>>>>> As you may know, I have simulcast and archived AfPIF every year since
>>>>> 2013.
>>>>> https://isoc.live/?s=AfPIF
>>>>>
>>>>> This was previously funded by the Internet Society. After they
>>>>> withdrew funding for ISOC.LIVE last year, Vint Cerf personally covered a
>>>>> further few months out of his own pocket, just covering AfPIF 2024,  and I
>>>>> was able to also include EN/FR captions nd transcripts of all sessions.
>>>>> https://isoc.live/18269/
>>>>>
>>>>> This year, however, I am currently without funding to continue the
>>>>> tradition.
>>>>>
>>>>> It is possible an ISOC Chapter Chapter could apply for a "Beyond The
>>>>> Net' grant to cover, but there is an 8 week lead time for such, so time
>>>>> grows short.
>>>>>
>>>>> Another option could be a sponsor, and they could be promised to have
>>>>> their name on screen at all times
>>>>>
>>>>> I would price it at $1k/day. I have added AI summaries to my workflow.
>>>>>
>>>>> Any thoughts?
>>>>>
>>>>> J
>>>>>
>>>>> --
>>>>>
>>>>>
>>>>> Joly MacFie
>>>>> Whatsapp +12185659365
>>>>>
>>>>
>
> --
> Cell / WhatsApp / Signal: +256786350675
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.afpif.org/pipermail/afpif/attachments/20250827/67123c96/attachment-0001.htm>