Eddie AI
A speaker presenting in front of a huge screen of glowing diagrams to a developer audience.

How to Edit a Conference Keynote or Talk Video

A conference produces dozens of talks, and each one needs the same edit: trimmed, synced, slides visible, speaker named, captioned. The last step is where many recordings lose their audience. Automatic captions garble names and technical terms, and an upload called "Day 1 Stage A part 3" does not match what people type when they look for the talk. The worked example follows one talk, Raj Patel's 28-minute "Scaling search to a billion queries", from the full-day recording to a YouTube upload.

The short answer

Trim each talk from the speaker's first line to the end of their last answer in the Q&A, show the speaker and the slides together, and put the talk title and the speaker's name on screen. Before you publish, fix names and jargon in the transcript, upload your own caption file instead of relying on YouTube's automatic captions, and use the talk title from the programme as the video title. Eddie does the edit from prompts, and you reuse the same prompts for every talk.

The deliverables

What you will end up with

  • Each talk trimmed from the speaker's first line to their last answer.
  • The speaker and the slides visible together, in a layout that fits the talk.
  • A title card with the speaker's name and the talk title, and a transcript with those names spelled right.
  • An SRT caption file and a YouTube title that match the talk.
  • Two or three short clips per talk for social.

6 steps

Step by step

  1. 01Import the day and sync the stage
  2. 02Split the day into talks
  3. 03Lay out speaker and slides
  4. 04Add the title card and fix the names
  5. 05Pull clips and export
  6. 06Publish on YouTube with the talk's title and your captions
  1. Step 1

    Import the day and sync the stage

    Upload the full-day recordings from the stage as separate files: the cameras, the slide capture and the audio desk recording. Sync them once for the whole day.

    What you type
    “Sync the wide camera, the speaker camera, the slide capture and the room audio as one multicam group, using the room audio for every cut.”

    What Eddie does

    • Transcribes the full day and syncs the files by sound into one group.
    • Sets the desk recording as the group's audio.

    What you get: A whole day of talks, synced, in one project.

    Eddie project with a synthetic full day of a conference stage: a wide camera, a speaker camera, a slide capture and room audio.
  2. Step 2

    Split the day into talks

    Ask Eddie to find each talk's start and end from the transcript: the host's introduction, the speaker's first line, the last answer in the Q&A.

    What you type
    “Find where each talk starts and ends: from the speaker's first line after the host introduces them to the end of their last answer in the Q&A. List them with the speaker names.”

    What Eddie does

    • Searches the transcript for each host introduction and each talk's last answer.
    • Returns each talk with its start, end and speaker.

    What you get: The day split into talks.

    List of conference talks found in a full-day recording with start times, speakers and lengths.
  3. Step 3

    Lay out speaker and slides

    Viewers need the slides. Show the slides and the speaker side by side for the whole talk, with the slides taking the larger share of the frame so code and diagrams stay readable. If a talk is mostly code or diagrams, make its edit slides only, with the speaker's audio.

    What you type
    “Make an edit of Raj Patel's talk with the slides on the left two thirds and the speaker camera on the right third, and remove the 40-second pause while the clicker failed.”

    What Eddie does

    • Builds the talk edit and places the slide capture and the speaker camera as layers on the canvas.
    • Cuts the 40-second pause out of the talk.

    What you get: A talk viewers can follow on a laptop or phone. Without the pause, the 28-minute talk runs 27:20.

    Talk timeline with one slides-and-speaker layout for the whole talk, and the slide capture and speaker camera layers cut where the clicker pause was removed.
  4. Step 4

    Add the title card and fix the names

    Put the speaker's name and the talk title on screen at the start. Then correct the transcript wherever speech recognition misheard a name, a company or a product term: the captions and the SRT file are built from it. Leave burned-in captions off the full talk if it is going to YouTube. You will upload them as a file in the last step, so they are not drawn twice when a viewer turns captions on.

    What you type
    “Add a title card for the first 5 seconds: "Scaling search to a billion queries, Raj Patel, Northwind". Then correct the transcript: "North wind" should be "Northwind" and "Rush Patel" should be "Raj Patel".”

    What Eddie does

    • Places the title on the edit's title track for the first 5 seconds.
    • Replaces each misheard phrase everywhere it occurs in the source transcript, keeping every word timing, and rebuilds caption tracks built from that transcript.

    What you get: A talk that names its speaker, with a transcript you can trust for captions.

    Checklist for a finished talk: title card for the first five seconds, corrected names in the transcript, a removed pause and the desk audio.
  5. Step 5

    Pull clips and export

    Two or three short clips per talk promote the recording. Social clips get their captions burned into the MP4. Export the full talk as an MP4 for YouTube or your website. The cloud MP4 render takes edits up to 60 minutes, so Raj's 27:20 talk and the 41-minute keynote fit; a longer session needs a timeline export or the desktop app.

    What you type
    “Make 2 vertical clips of 45 seconds from the strongest moments of Raj's talk, on the speaker camera with social captions. Export the talk and the clips as MP4.”

    What Eddie does

    • Picks self-contained moments and builds vertical edits on the speaker camera.
    • Renders the talk and the clips in the cloud.

    What you get: The talk and its social clips, ready to publish.

    Eddie export menu with the MP4 cloud render selected, plus timeline formats.
  6. Step 6

    Publish on YouTube with the talk's title and your captions

    Use the talk title from the programme as the video title, so the upload matches what attendees search for. In YouTube Studio, choose Content, click the video, edit the title and select Save; titles are limited to 100 characters. Then upload the SRT: Subtitles, the video, ADD LANGUAGE if needed, ADD under Subtitles, Upload file, With timing, Continue, choose the file, Save.

    What you type
    “Export the transcript of Raj's talk edit as SRT, and list the title card text so I can paste it into the YouTube title.”

    What Eddie does

    • Writes the SRT from the transcript of the edit, timed to the cut you made rather than to the raw footage. This export costs no credits.
    • Returns the title card text exactly as it sits on the edit's title track.

    What you get: A YouTube upload titled like the talk, with captions that spell every name right.

    Checklist for publishing a talk on YouTube: the video title copied from the talk title, the YouTube Studio path for uploading an SRT, and the path for editing automatic captions.

Craft

What makes it work

Record the slides directly
A capture of the presentation laptop's output is far sharper than a camera pointed at the screen. Ask the AV team for it.
Sync once for the whole day
Group the day's files once, then cut each talk from the group. Every talk uses the same sync and audio.
Give the slides the larger share
Keep one side-by-side layout for the whole talk and give the slides most of the frame, so small text stays readable. For a talk that is mostly code, publish a slides-only edit with the speaker's audio instead.
Fix names before you export captions
Correct the speaker and company names in the transcript first. A caption file you already downloaded does not change when you fix the transcript, so export the SRT again after any correction.
Fix automatic captions you already published
If the talk went live with YouTube's automatic captions, open YouTube Studio, choose Subtitles and the video, select Edit next to the language, then Duplicate and edit and Continue. Click any line to change its text and select Publish. YouTube makes a new caption track with your changes.
Title it the way people search
YouTube search weighs how well the title, tags, description and video match the search. Lead with the talk title as printed in the programme, then the speaker and the event, all within 100 characters.
Reuse the prompts for every talk
Once one talk is right, send the same layout, title card and clip prompts for the next one, changing only the speaker, the talk title and the times.

FAQ

Questions

Can Eddie split a full-day recording into separate talks?
Yes. Eddie finds each talk's start and end from the transcript, and you make one edit per talk.
Can I show the slides and the speaker together?
Yes. Eddie can place several video layers on the canvas, such as the slides and the speaker side by side.
Why are YouTube's automatic captions wrong on my talk?
They come from speech recognition, and YouTube says they can misrepresent speech because of mispronunciations, accents, dialects or background noise. Names, company names and technical terms suffer most. They are also made only in the video's default language and may not be ready straight after upload. Upload your own SRT instead, or edit the automatic track in YouTube Studio.
What file does YouTube need for captions?
SubRip (.srt) is one of YouTube's basic caption formats. It must be plain UTF-8, and YouTube ignores styling. Choose With timing when you upload a file that already has timestamps, as an SRT from Eddie does.
Should the YouTube title match the talk title?
Yes. Attendees search for the title they saw in the programme, and YouTube search looks at how well the title matches the query. Use the talk title first, then the speaker and the event, for example "Scaling search to a billion queries | Raj Patel, Northwind | DevConf".
What about panels with several speakers?
Sync a camera per panellist if you have them, and Eddie can switch to whoever is speaking. Name the speakers so captions and titles are right.
Can I remove technical problems from a talk?
Yes. Name the problem and roughly where it happens, such as a clicker failure, and Eddie removes that stretch.
Can I publish the talks with the slides only?
Yes. Ask for the slide capture full screen with the speaker's audio, and Eddie builds the talk that way. It suits technical talks where the slides carry most of the content.

Edit your conference talks with Eddie

Sign up, add your footage and type the first prompt above. New accounts start with free credits.