Turn an Interview Into Two Clips With Riverside Agents

• Create a folder named Riverside Clip Practice on your computer.
• Open the interview’s file page and choose Original file.
• Save it as interview-source.webm, then save the credit below as source-credit.txt.
• Play several sections and confirm you hear speech and can count the speakers.

Credit for the finished excerpts:

Source: "Interview with Steve Wozniak" by ConversationEDU, dated October 11, 2016 in the source record.
Original: https://www.youtube.com/watch?v=920QAxtook8
Download and license record: https://commons.wikimedia.org/wiki/File:Interview_with_Steve_Wozniak.webm
License: Creative Commons Attribution 3.0 Unported (CC BY 3.0).
https://creativecommons.org/licenses/by/3.0/
Changes in these excerpts: shortened, reframed vertically, and captioned.
Independent editing exercise. No endorsement by the speakers or original publisher is implied.

The file page lists a 14-minute, 48-second interview at 1920 × 1080, dated 2016. Its CC BY 3.0 license allows sharing and editing with credit, a license link, and notice of changes. Keep the historical date visible. Use the excerpts for editorial practice, without implying an endorsement.

• You saved a webpage or thumbnail: Choose Original file on the file page, then save the video itself. The listed original is about 160 MB.
• Your player cannot open WebM: Use the optional MP4 conversion in Step 2, then check the converted file before uploading.

• Open interview-source.webm in a player that supports WebM.
• Check the beginning, middle, and end for picture and speech.
• Confirm a duration of roughly 14:48 and a 1920 × 1080 landscape frame.
• Count two speakers: the interviewer and Steve Wozniak. Both appear together in some shots.
• Keep the original file and source-credit.txt together. Upload this WebM in Step 3.

Riverside lists WebM with VP8 among its supported video formats. The original file uploaded and produced an editable video and transcript in this walkthrough. Keep the full interview so the agent can find passages in context.

• Playback or upload fails: Open the file in HandBrake, choose General → Fast 1080p30, and export the full duration as interview-source.mp4. Check MP4, H.264 video, and AAC audio. The preset documentation describes these settings. This is a fallback, not a required installation.
• The converted file is silent or short: Select the source audio track in HandBrake’s Audio tab and encode the whole title, not a chapter or selected range. Play the output before uploading. If you use this fallback, replace the source filename in the selection prompt with interview-source.mp4.

• Open Riverside’s editing workspace and confirm your plan supports external uploads. We did not find a used-upload-hours meter in this walkthrough, so check the plan’s allowance before uploading.
• Click Upload → Select file and choose interview-source.webm.
• Set 2 speakers; leave Separate video tracks off.
• Click Upload and keep the page open until the upload finishes.
• In the editor, name the full-length edit Full interview source review. Compare speech with transcript passages near the beginning, middle, and end.

The upload flow creates a transcript for an external recording. Leaving video separation off preserves the source composition. This walkthrough fits that single frame into a portrait canvas without splitting or switching speakers. Step 5 uses the Grid/Fit control inside Smart layouts to preserve that frame.

Riverside may offer an automatic editing plan or create Magic Clips after upload. Do not click Start editing on a plan that removes filler words, shortens pauses, enhances audio, or adds visuals. We will request only the changes this exercise needs.

• Riverside rejects the file: Use the MP4 fallback in Step 2. Convert the video; do not merely rename the WebM extension.
• The transcript uses the wrong language: Set Settings → Studio settings → General → Transcript & captions language to English, then upload again. The setting affects future transcriptions.

• Select Clipper in the agent sidebar and paste the selection prompt below.
• Read its candidate ranges, or preview the clip cards if it creates them. A suggestion is not necessarily a saved edit.
• Watch the source just before and after each proposed excerpt.
• Keep two complete passages, each 30–60 seconds, covering different ideas.
• Open Editor directly and request one separate draft at a time. Open the new draft before applying its trim; keep Full interview source review unchanged.
• Check the draft’s actual duration and first and last words. Review any large-trim confirmation before applying it. Do not continue with a partial word or sentence at either edge.

Clip selection prompt:

Use only interview-source.webm, the uploaded 2016 ConversationEDU interview. This is an independent editing exercise for a professional audience interested in building products and businesses.

Create three candidate clips, aiming for 30 to 60 seconds each. Choose different, self-contained ideas that the speaker actually discusses. Do not force a topic the interview does not cover.

Each candidate must use one continuous passage in the original order. Start and end at complete thoughts. Include the question when an answer needs it to make sense. Preserve qualifications, dates, negatives, and any context that affects the meaning.

For each candidate, show its original start and end timestamps, first and last spoken sentence, and a one-sentence reason to keep it. If you cannot produce three suitable candidates, explain the limit rather than inventing words or timestamps.

Keep the original recording intact. Do not add music, B-roll, generated visuals, rewritten speech, voice cloning, dubbing, eye-contact correction, or internal sentence rearrangement. Do not apply a saved brand kit to this archival interview.

Keep these as drafts. Do not connect an account, publish, schedule, or send anything. Show any credit charge before running a charged action.

The walkthrough used these two continuous source passages. Timestamps are starting points to verify against your transcript, not a substitute for checking the words:

Create the first draft in Editor’s own chat:

Create exactly one new draft edit named Interview clip 01 from the original full-length uploaded interview. Keep Full interview source review and the original recording unchanged.

The intended continuous source range is 04:42.57–05:28.14 (282.57–328.14 seconds), about 46 seconds. It starts “And I knew that for a home computer…” and ends with the complete sentence about games never having been software before. Exclude the following sentence beginning “When they were hardware…”.

Create the separate draft first and offer to open it. Do not run your suggested automatic editing plan. Make no internal cuts, remove no filler words or pauses, change no audio or framing, and add no media. Do not export or publish.

After opening Interview clip 01, apply its trim:

Work only on the open draft Interview clip 01. Keep the single continuous original-source interval 04:42.57–05:28.14 (282.57–328.14 seconds). Apply the two cuts outside that interval; preserve every word, pause, and frame within it. Include the complete opening “And I knew that for a home computer…” and end after the sentence about games never having been software before. Exclude the following word “When” and its sentence.

Keep the original recording, Full interview source review, and every other edit unchanged. Add no internal cuts, enhancements, layouts, generated media, music, export, or publication.

Show any large-trim confirmation for this duplicate. After applying it, report the actual timeline duration, first and last complete sentences, and any uncertain boundary. Do not describe a proposed trim as completed.

Create the second draft separately:

Create exactly one new draft edit named Interview clip 02 from the original full-length uploaded interview. Keep Full interview source review, Interview clip 01, and the original recording unchanged.

The intended continuous source range is 03:31.37–04:17.36 (211.37–257.36 seconds), about 46 seconds. Include the full question beginning “So there’s been much talk about disruption of the higher education sector…” and the answer through “At least computers don’t do it yet.” Preserve the speaker’s distinction between university-level and younger students.

Create the separate draft first and offer to open it. Do not run your suggested automatic editing plan. Make no internal cuts, remove no filler words or pauses, change no audio or framing, and add no media. Do not export or publish.

After opening Interview clip 02, apply its own trim:

Work only on the open draft Interview clip 02. Keep the single continuous original-source interval 03:31.37–04:17.36 (211.37–257.36 seconds). Apply the two cuts outside that interval; preserve every word, pause, and frame within it. Include the complete question beginning “So there’s been much talk about disruption of the higher education sector…” and the answer through “At least computers don’t do it yet.” Use whole-word edges.

Keep the original recording, Full interview source review, Interview clip 01, and every other edit unchanged. Add no internal cuts, enhancements, layouts, generated media, music, export, or publication.

Show any large-trim confirmation for this duplicate. After applying it, report the actual timeline duration, first and last complete sentences, and any uncertain boundary. Do not describe a proposed trim as completed.

Clipper suggests and edits clips, but you still need to check its choices. One suggestion in our test ran over 60 seconds and ended mid-sentence, so we dropped it. An opening such as “That’s why it worked” may need the sentence before it.

Ask Editor directly to create each draft. Asking Clipper to pass the job to Editor did not work in our test. Asking for two drafts at once also failed to produce two separate edits. Create one, open it, and check its length before starting the next.

Cutting a 14:48 copy down to about 46 seconds removes roughly 95% of it. Riverside may ask you to confirm this more than once. Check the edit name and the part you want to keep each time. Then check the timeline: the chat may say it has made a cut before the edit has changed.

• Clipper returns too many clips or an unsuitable passage: Select a complete 30–60-second passage in the source transcript and click the clapper icon to create a new edit.
• The agent does not load: Save useful notes, clear the current agent chat, and refresh the page, as Riverside recommends.

• Open Interview clip 01. Click the aspect-ratio control above the preview and choose 9:16 Vertical.
• Open AI tools and the settings control beside Smart layouts.
• Keep Grid as the style. Under Video frame, select Fit, not Fill. In this test, selecting Fit enabled Smart layouts; with one unsplit video track, it retained the complete landscape frame.
• Check that the picture sits across the middle with plain dark space above and below. Keep the original Conversation source mark visible.
• Inspect the beginning, middle, and end for cropping or stretching. Repeat these controls on Interview clip 02.
• Leave pause removal, Magic Audio, Smart Mute, eye-contact correction, video enhancement, and AI color off. Add no music, B-roll, generated media, or camera motion.

Editor said it could not make this layout vertical. The control above the preview worked, so use that instead. Fit keeps the whole picture in view. The dark space gives the captions and source label room without covering the interview.

• The canvas stays wide: Click the aspect-ratio control above the preview and choose 9:16 Vertical.
• The picture is cropped: Confirm Video frame → Fit, not Fill. Check the result after the preview finishes updating; do not rely on the agent’s description of the layout.

• Open Captions and choose DM Mono for the first clip.
• Select the caption box, set font size 64, keep 2 lines and a dark background, and choose No animation. Drag the box into the dark area below the landscape frame; apply to All scenes.
• Listen throughout; select each transcript error, click Correct, and enter what you hear.
• Choose Text → Body, replace the placeholder with the label below, and use white Inter Regular, size 64. Position it above the landscape picture. Zoom out the timeline and drag the label bar’s left and right edges to cover the entire clip, not just its default five seconds.
• Repeat on the second clip, checking captions and labels at a phone-sized preview.

On-screen source label:

From a 2016 interview
ConversationEDU · CC BY 3.0

Caption controls set text style and position. Place captions below the landscape picture, away from the canvas’s bottom edge. Keep the source label above it. Drag both away from any existing source mark.

Use Correct for transcription mistakes. These corrections also update captions and downloadable transcripts in other edits from the same recording. Check the exact phrase before saving; do not use a global replacement when the word is correct elsewhere. Deleting words from the editing transcript can cut the recording. Keep audio and video unchanged in this step; use no AI voice replacement or VideoDub. The short label accompanies the full credit in your post text.

• Captions cover a face or source mark: Select the caption box and drag it into the empty area. Check all scenes.
• The label vanishes partway through: Extend the text overlay’s duration to the clip’s end on the timeline.

• Select Marketer in the agent sidebar with the two finished edits in the current project.
• Send the post-drafting prompt below without connecting a social account.
• Check every claim against the matching clip and keep the 2016 context.
• Save each complete draft as interview-clip-01-post.txt or interview-clip-02-post.txt.

Post-drafting prompt:

Read the finished edits Interview clip 01 and Interview clip 02 in this project. Use their actual spoken content to draft one LinkedIn post for each.

Start each post with: From a 2016 interview:
Then give a plain, accurate summary of the clip in no more than 60 words, excluding the source credit. Use no quotation marks unless you have checked the exact words against the clip. Do not invent an outcome, turn old comments into current news, or imply that the speakers endorse Riverside or The Rundown University.

Append the following source credit to each post without shortening its URLs:

Source: "Interview with Steve Wozniak" by ConversationEDU, dated October 11, 2016 in the source record.
Original: https://www.youtube.com/watch?v=920QAxtook8
Download and license record: https://commons.wikimedia.org/wiki/File:Interview_with_Steve_Wozniak.webm
License: Creative Commons Attribution 3.0 Unported (CC BY 3.0).
https://creativecommons.org/licenses/by/3.0/
Changes in these excerpts: shortened, reframed vertically, and captioned.
Independent editing exercise. No endorsement by the speakers or original publisher is implied.

Return both drafts here, labeled with their matching edit names. If you cannot read an edit, identify the gap and stop. Do not rely on another recording or previous studio conversation.

Do not generate a thumbnail, connect a social account, publish, schedule, send, or change the videos.

The Marketer agent can write posts as well as publish them. Stop at drafts here. Keep both credit URLs with each excerpt wherever you later have permission to share it.

• The post turns an old remark into current news: Keep “From a 2016 interview” and remove claims the clip cannot support.
• The agent requests social access: Decline it. Ask for text in the conversation; this exercise needs no publishing connection.

• Open Interview clip 01 and select Export → Download → Video → 1080p.
• Enable Remove watermark if your plan permits it. Leave Normalize audio and Remove background noise off, then click Download video.
• Repeat for Interview clip 02 and save both videos beside their post files.
• Open each MP4; check portrait dimensions, 30–60-second duration, sound, captions, source label, and complete sentences.
• Confirm the folder contains the five deliverables below and no clip makes an unsupported claim.

Final filenames:

interview-clip-01.mp4
interview-clip-01-post.txt
interview-clip-02.mp4
interview-clip-02-post.txt
source-credit.txt

Riverside exports clips as MP4. For this 9:16 export, check for 1080 × 1920 pixels in the downloaded file. The portrait canvas does not add detail to the original footage. Leave enhancement and noise-removal options off unless you have listened to and approved their effect.

Keep the Riverside drafts and original source for later changes. Your handoff now contains two captioned excerpts and their matched post copy. Nothing needs to go live during the exercise.

• Export offers a lower resolution or keeps a watermark: Check your plan’s export rights. Use a clearly labeled practice preview or resolve access before delivering the final set.
• Riverside says the download will start, but no file arrives: Open Exports in the project sidebar. Open the finished video and click Download export. That worked in our test. Check there before exporting the same clip again.
• A download has the wrong words or cut points: Return to that edit, correct it, and export again. Check the downloaded replacement rather than only the preview.

Watch the two downloaded videos below. The first runs 45.58 seconds; the second runs 46 seconds. Both are 1080 × 1920, with captions included in the picture.

Source: Interview with Steve Wozniak, ConversationEDU, 2016. CC BY 3.0. Shortened, reframed, and captioned. No endorsement implied.

A note on our check: We checked the files, pictures, and machine transcripts, but could not listen to the audio. Before sharing these example clips, listen to both in full. Two short phrases in the first clip still need that check: the word before “modifying” at about 14–17 seconds, and the contraction in the final sentence at about 43–45 seconds.