Create Short Video Clips from Tella Transcripts with Claude

Watch: Create Short Video Clips from Tella Transcripts with Claude

A long-form video can contain several useful clips. The hard part is finding clean boundaries and preserving enough context.

Disclosure: This article contains a Tella affiliate link. Purchases through it may earn me a commission.

Tella's MCP server lets a compatible assistant access your workspace through natural language. MCP is a standard way for an assistant to call tools and receive data. Tella documents tools for finding videos, reading transcripts, and creating duplicates in its MCP documentation.

This workflow helps you identify candidate moments and create trimmed copies. Review every clip before sharing it. A transcript can contain errors, and a strong sentence can still need context.

Prerequisites: Connect Tella to your assistant first. See Part 1: Getting Started with Tella MCP. Part 2 covers transcript-to-blog workflows.

The three-prompt workflow

Use three prompts for the core process:

1. Retrieve the video and its transcript.
2. Ask for self-contained clip opportunities.
3. Duplicate the source with selected start and end times.

Review suggestions before creating anything.

1. Retrieve the video

Start with a specific title when possible:

Use Tella to find the video titled "[VIDEO TITLE]".
Return its ID, duration, chapters, and transcript when available.
Do not edit, export, or delete anything.

If you do not know the title, search for a distinctive phrase:

Search my Tella library for videos about "[TOPIC OR PHRASE]".
Search titles and transcripts if both are available.
Return the matching titles and IDs.

Tella documents list_videos, search_videos, and get_video for these actions. get_video can include transcript and chapter data. Exact tool names depend on your assistant. Confirm the returned video before continuing.

2. Find clip opportunities

Tell the assistant what each clip must accomplish. Do not treat a fixed duration as a platform rule. Platform limits and recommendations change. Check the destination's current guidance, then provide a target range.

Using the transcript from this video, find three self-contained clip opportunities.
Target duration: [TARGET RANGE IN SECONDS].
Audience: [WHO WILL WATCH].
Purpose: [TEACH, TEASE, OR EXPLAIN].
Each clip must make sense without the full video.
For each suggestion, return the start and end timestamps, main idea,
reason it works alone, suggested title, and missing context.
Do not create clips yet.

Ask for a shorter or longer range based on the destination and idea. These are editorial choices, not universal platform requirements. Check each suggestion against the transcript. Remove clips that start mid-sentence or end before the conclusion.

Tella transcript and timeline responses may expose times in milliseconds. The trimming tool uses a different contract. Convert the boundaries before creating the clip.

3. Create the trimmed clip

Choose one suggestion and state its boundaries clearly:

Use Clip 1 from the suggestions.
Create a duplicate of the Tella video.
Trim it with duplicate_video using startTime [START IN SECONDS]
and endTime [END IN SECONDS].
Name the copy "[CLIP TITLE]".
Ask for confirmation before any other edit.
Return the new video ID and link when available.

Tella's current video tools reference defines duplicate_video with videoId, optional name, startTime, and endTime. The trim values are seconds. It also supports chapterIndex for extracting a chapter instead of a time range. Do not pass millisecond values to startTime or endTime.

When the assistant requests permission, check the operation and target video. Approve only the duplicate you intend to create. Then open the new video and check its first and last seconds.

For vertical or square destinations, specify the desired canvas separately. Trimming does not change the aspect ratio. Tella documents presets such as 1080x1920 and 1080x1080. Verify current settings and destination requirements before exporting.

Troubleshooting

If Tella cannot find the video, search by title or a distinctive transcript phrase. Confirm the video ID before requesting details.

If the transcript is missing, check whether transcription is still processing. Tella reports transcript details as unavailable while processing, after a failure, or when the transcript cannot be read. Retry after it becomes ready.

If a suggestion starts or ends poorly, ask for new boundaries. Include the nearest complete sentence and required setup.

Final review

Before sharing a clip, check:

• The clip uses the intended source video.
• The opening and ending work without extra context.
• Names, numbers, and product terms are correct.
• The duration matches the destination's current requirements.
• The aspect ratio and captions suit the destination.
• The title and link match the clip.

Human review decides whether the clip represents the original idea accurately.

Next step

Pick one video and ask for three candidate moments. Review their boundaries beside the transcript. Create one duplicate only after the candidate passes that check.

Convert one boundary example

Suppose the transcript boundary is 90,000ms to 125,000ms. Divide each value by 1,000. For duplicate_video, use startTime: 90 and endTime: 125. The resulting range is 35 seconds. Tella's video tools reference defines those trim parameters in seconds. Its chapterIndex option uses a zero-based chapter index and cannot accompany the time range. Open the copied clip and review both boundaries before exporting it.