Claude can send a YouTube recording to Prolifik's Clipper and retrieve the finished clips when processing completes. The most useful request says which conversation or lesson to preserve, not just how many Shorts you want.
Start with a finished recording you own or have permission to repurpose. Give the assistant its actual source URL and a reason for the clips. A title or thumbnail alone does not establish what the speakers said, and an assistant's written summary is not the rendered footage.
This walkthrough uses a fictional founder interview about replacing manual appointment reminders. The sample requests are illustrative. They do not represent a clipping test on a real account.
Define a clip that can stand on its own
Our fictional interview contains a useful passage: the founder explains why the team first improved reminder wording before changing the delivery system. A short excerpt should preserve that sequence. Cutting only the line “we automated the reminders” could imply a different lesson.
Give Claude this selection brief:
From my supplied interview URL, request three candidate vertical clips about appointment reminders. Favor complete explanations with their necessary setup. Preserve qualifications about when manual reminders still make sense. Do not select a sentence simply because it sounds like a universal promise. Return drafts for download.
These are selection instructions, not guarantees about the output. Every returned clip still needs inspection. If the source does not contain three useful standalone moments, a smaller usable set is better than stretching a weak passage to meet a requested count.
Check the source and the clipping allowance
Connect Prolifik through the current MCP connection instructions, then enable the connection in the Claude environment you use. Claude account and workspace permissions affect access. The content and clipping tools are available through Prolifik's Claude connection; this workflow does not require its separate AI Studio or cloning features.
Ask for get_clip_minutes before submitting the job. Clipping uses the account's source-minute allowance, so the recording's length matters even if the final clips will be short. Read the returned allowance instead of estimating entitlement from an old pricing table.
Confirm that the URL points to the intended finished recording. If the source is unavailable to the service, correct the source or use a supported upload path. Repeatedly submitting the same inaccessible link will not produce different footage.
Specify framing and useful constraints
The current create_clip_job schema includes the source URL, requested clip count, duration range, aspect ratio, caption style and selection guidance. It also exposes options for the type of recording and speaker-related treatment.
For the illustrative interview, a request can say:
Use the supplied owned YouTube interview. Request three clips between 30 and 60 seconds, in 9:16, with readable captions and speaker framing suitable for an interview. Use the selection brief above as the job's guidance. Keep the original job identifier and use one idempotency key for this attempt. Do not publish the clips.
The duration range is a creative preference for this example, not a platform rule or a quality promise. If the useful answer takes longer, review whether to request a different range. Do not force a clip to end before the speaker finishes the point merely to keep it under a chosen number.
The podcast clipping guide covers how interview structure affects selection. For multi-speaker footage, inspect the crop whenever the active speaker changes.
Follow the job without starting it again
The creation call queues processing. Claude should retain the returned job identifier, use the available status lookup and retrieve the completed job through get_clip_job. Repeating the original creation request is not the right way to ask whether processing has finished.
If the first response is interrupted, recover the existing request using its identifier or idempotency key. An uncertain response does not prove that the job was rejected. Keep the source URL and request together so the assistant can distinguish this interview from another recording with a similar title.
A useful progress report states the actual job status. A useful completion report lists the returned clip identifiers and media. If processing fails, preserve the failure reason and decide what must change before another attempt.
Inspect the first and last thought
Play each clip from its first frame. In the reminder interview, the viewer should know what the founder is discussing before hearing “that was the problem.” Pronouns and references that made sense in the full interview can become confusing after extraction.
Watch the ending too. A clip that stops after “but only when” reverses or hides the speaker's qualification. Captions should match the speech, including product names and numbers. Automated transcription can make mistakes even when the sentence sounds plausible.
Ask the reviewer to record concrete decisions: “Candidate two keeps the full explanation but spells the product name incorrectly”; “candidate three starts after the question it answers.” The clip start and ending guide provides a more focused boundary review.
Keep the usable output with its source
Save the accepted clips with their source and job identifiers. That record lets a teammate retrieve the correct media and understand why an excerpt was selected. Signed media URLs are temporary, so preserve identifiers as well as the downloaded files.
If a clip needs exact re-trimming, use the dashboard workflow; Prolifik's MCP documentation does not expose that editing operation. Do not ask Claude to claim it changed a boundary through a tool that has no such control.
Once the excerpt and captions work, use Prolifik Clipper for the source-to-short workflow or make a separate publishing request to a selected connected account. The successful outcome is an understandable clip from the real recording, with a playable file and known provenance.
Turn your recording into useful clips
Use the source you own, preserve the complete thought and inspect the returned short videos.
Open Prolifik ClipperProduct and primary-source checks: 15 October 2026. Worked examples are illustrative unless a real source is identified.