Documentary video editor

Yes. Analysis transcribes every interview word by word in 90+ languages, labels who is speaking, sorts each file into topics and subtopics, and writes a short description of every clip, so B-roll is searchable by what's on screen.

It takes about 20 minutes per hour of video, counted file by file, and you can keep working while it runs. A project has no limit on total footage, and each file can be up to 10 hours long.

Check names and terms lists the names and places analysis found, and one correction fixes every clip and draft in the project.

Yes. Attach the paper cut to chat and Selects matches each line to the take it came from, then cuts them in your order.

1. Export the transcripts from Export > Transcript as a text file, with timecodes, speaker names and chapter titles.
2. Write the paper cut in your own document.
3. Attach it to chat as a .docx, .pdf, .txt or .md file under 45,000 characters, and ask for the cut.

If a subject said a line twice, ask chat to keep the best take. Scanned PDFs can't be read, so attach a file with real text in it.

Yes. Selects syncs 10 or more camera angles plus separate lav, boom and wireless audio, by audio waveform or by timecode, and matches each voice to its mic.

Then ask chat to switch cameras, with a rule if you have one, like stay on the close-up while she answers and go wide between questions. Waveform sync needs sound that overlaps, so a camera that recorded no audio is placed by hand.

Yes. Describe it in Generate video, like a slow drone pass over the coastline at dusk, and Selects generates it, up to 30 seconds a clip on Seedance 2.5. It lands in your draft at the playhead and in the project next to your footage.

Generation runs on credits, and the cost is shown before you generate. It is generated, not filmed, so label it wherever your film's standards ask you to.

Yes. Paste the narration script into Generate audio and pick a voice, or ask chat for a voiceover. ElevenLabs v3 suits narration and reads up to 5,000 characters a run.

Seed Audio 1.0 can match a voice from up to three short reference clips. The voice speaks the language the script is written in, and you swap in the real read once it's recorded.

It doesn't have to. Let Selects do the logging, the stringouts and the first assembly, then hand the cut to Adobe Premiere (.prproj), Final Cut Pro (.fcpxml), DaVinci Resolve (.drp) or Avid Media Composer (.aaf) to finish.

Edit points, tracks, speaker labels and links to your original media come across. Premiere mirrors your Selects folders as bins, and Premiere and DaVinci Resolve can bring topic titles in as markers.

Effects, color, text and audio levels don't. Transitions arrive as plain cuts and graphics as gaps, so the grade, the mix and the titles happen in your editor.