ChabadLabs AI FOR SHLICHUS
EN

Video Creation

Reels from event footage, clips from shiurim, and AI-generated video: what shluchim use in late 2026, now that agents can do the cutting.

On this page

"How do I make a reel from this 90-minute class?" and "how do I turn 40 phone clips from the event into one video?" are two of the most-asked questions in the group. In May 2026 a member called short-video editing one of the most requested AI tasks among shluchim. By October 2026 the answer had changed: AI agents can now watch the footage, pick the moments and do the cutting, with a person giving notes.

This page is the map. The tool-by-tool list, with prices, is on Video Editing Tools.

Event footage into a reel

The simple route. CapCut's AutoCut builds a short recap from your clips and photos; it was the standard answer for a 90-second event reel in March 2026, and for simple reels the advice in August 2026 was "stick with CapCut."

The agent route (new in fall 2026). In October 2026 a member had Claude Opus 5.5 edit an event video. It went through 73 clips (about 21 minutes), sent the best moments for approval first, picked 26 shots for a 44-second video, cut on the beat, pulled the logo from Canva, and took 9 rounds of notes. To teach it the house style, the video editor screen-recorded himself editing while talking through his choices, and Claude turned that recording into editing rules. Another member with 2,500 photos had Claude Desktop pick the best ones and Claude Cowork assemble the reel. Members also report that the Tab agent made "a decent" highlight reel from a Google Photos album.

What no tool fixes. Footage that was not shot for a reel (wide, static, long) will not become an engaging reel. Film short, close, vertical clips at the event if you want a good result (members, October 2026).

A shiur into short clips

  1. Get a transcript with timestamps. See Transcription.
  2. Let AI pick the clips. Paste the transcript into Claude or ChatGPT:
Here is a transcript of a shiur with timestamps. Find the 8 to 12 passages that stand on their own as 30- to 60-second clips: a complete thought, a story, or a sharp question and answer. For each, give the start and end timestamps, a one-line hook for the caption, and why it works without context. Do not change any wording of the quotes.
  1. Cut. Opus Clip and CapCut both let you cut by deleting words from the transcript. Since September 2026 members have also had Claude Cowork download the video and cut excerpts at the timestamps; the first try mismatched the file format and then fixed itself.
  2. Caption, then post.

In October 2026 one member had the Tab agent find a specific segment in a 3-hour radio interview and make portrait and landscape subtitled clips with a waveform in about 10 minutes, which replaced his paid Opus Clip plan. That is one person's result. Step by step: How do I turn a long shiur into short video clips?

Generating video with no footage

Sora shut down in April 2026. What members use instead:

  • Grok for short clips; recommended "for vids" in October 2026. Check the current plan, since free video generation has been limited.
  • Kling (kling.ai): Kling 3.0 (February 2026) makes clips up to 15 seconds with sound; Kling 4.0 is the current model on its site.
  • Google: Veo 3.1 and Gemini Omni, in the Gemini app and Google Flow with a Google AI subscription. In May 2026 a member recreated a choir-style video by having AI describe a real video in detail as JSON, then pasting that description into Gemini's video tool. It looked "100% AI," and members spotted errors such as peyos appearing mid-sentence.
  • Seedance 2.5 (ByteDance), also offered inside CapCut.
  • Multi-model sites such as Higgsfield and Lumeta, which put many image, video and voice models behind one account. In August 2026 a member shared a studio's video that brought shtetl paintings to life using Lumeta.

Every generated clip still needs the same care as an AI image: describe clothing, setting and action so the people look like your community. See Prompting AI Image Tools to Depict Your Community.

Talking-head and avatar video

  • HeyGen: free plan with 3 one-minute watermarked videos a month; Creator $29/month. Translation into 175 languages and dialects on paid plans.
  • Synthesia and D-ID, the Tel Aviv company members called "possibly better" (March 2026).
  • Google Vids in Google Workspace, with AI avatars and voiceovers in 24 languages. In May 2026 a member found the avatar flow records your voice by having you read numbers and does not accept an uploaded voice file.

A daily Torah video, start to finish

A member's October 2026 recipe for a short daily Chitas video: paste the Rashi or Chumash text into Claude and have it write the script; have ChatGPT read the script aloud as an MP3; let Claude assemble the video. The lesson he shared the next day: before narration, have ChatGPT rewrite the Hebrew words in the script with nekudos, and give it a list of Ashkenazi pronunciations. That fixed the Hebrew. More on voices: ElevenLabs.

For technical users

Two code-based video tools can be added to Claude Code as skills: Remotion (npx skills add remotion-dev/skills) and HeyGen's open-source HyperFrames (Apache 2.0, which turns HTML and animation into MP4). They were shared in July 2026 from a course promotion; nobody in the group has reported results.

Music and narration

  • Suno for songs and background niggunim.
  • ElevenLabs for narration in a stock or cloned voice.

Still unsolved in the group

  • Translating a shiur video with your own voice and lip sync. Asked in May 2026 without a tested answer. HeyGen and Captions now sell translation and lip-sync features; nobody has reported trying them on a shiur.
  • A searchable library of hours of footage (September 2026). Opus Clip was suggested; no one confirmed it works for this.
  • Automatic B-roll for a talking-head shiur. Captions and Opus Clip offer AI B-roll on paid plans; results in the group are unreported.

A word of caution

A rabbi in the group said in July 2026 that AI-made flyers, videos and posts, "even with the amazing prompt shared here, are usually very recognisable," and questioned whether it is the right use. Members still post AI video, but an obviously generated clip can cost more trust than it earns. See AI Flyer Aesthetics for the same problem in images.

Last updated October 2026 · Maintained by ChabadLabs AI Agent