AI analysis is what turns raw files into footage you can search, read, and edit. Selects transcribes every file, breaks it into topics, identifies who is speaking, and reads what is on screen.
Nothing is analyzed until you confirm. When you add files, Selects offers to analyze them, and you decide what runs. You are only charged for what finishes.
Start analysis
You choose what to analyze, and you can run it at any time, not just when the project is new.
Click [Analyze N clips] at the bottom of the [Project] panel. You can also hover any clip and click [Analyze], or select files, right-click them, and choose [Analyze N files].
[Analyze media for edit] opens with your clips listed. From [Analyze N clips], every clip starts checked. From one clip's [Analyze], only that clip starts checked. Check or uncheck clips to choose what gets analyzed.
The credit estimate updates as you change the selection, shown against your current balance. When it looks right, click [Analyze].
Files you have already analyzed are left out of the estimate and are not charged again. Audio in folders named Music, SFX, or BGM is left out of [Analyze N clips], so analyze those files one at a time with [Analyze] if you need their transcript. If the estimate is more than your balance, the dialog asks you to top up or upgrade before you can continue.
To analyze a finished file again, right-click it and choose [Re-analyze N files]. [Re-analyze media] opens with the cost. Re-analysis uses credits and replaces each file's transcript, including any edits you made to it.
Tip: If you shot multicam or recorded audio separately, sync those files into a synced timeline so the transcript follows the right audio. Sync and analysis can happen in either order. See Sync your files.
Note: A sync that came out wrong can be fixed after analysis. Because analysis runs on your individual files, correcting a sync does not throw the transcript away. The correction reaches every draft straight away, but clip positions and lengths, the cut structure, and topics and subtopics do not rebuild themselves. To update a synced timeline's speakers and topics, click [Update] on the [Update data] step in [Sync tools].
What happens while analysis runs
Selects works through each file in turn: preparing the audio, detecting silences, transcribing, dividing the transcript into topics, and finding key visual moments.
While it runs, the pill reads [Analyzing N clips] with a rough time remaining.
Click the pill to open the progress list. The run shows as one [Analyze files] row with a progress bar and how many files are done. Each file's own progress, or a warning icon if it failed, shows on its row in the [Project] panel.
You can keep working in Selects while analysis runs. Leave the app open so it keeps going. If you quit, analysis picks up again the next time you open Selects.
Tip: Leave [Keep this computer awake] switched on. It sits at the top of the progress list. If your computer sleeps partway through, analysis can fail. Your display can still turn off, but closing a laptop lid will still put the machine to sleep.
To stop a file that is running, hover its row in the [Project] panel and click [Cancel]. To stop the whole run, cancel it from its row in the progress list. Anything that already finished is kept, and you are not charged for it again.
What you get after analysis
Transcription. Every file gets its own transcript, with every word time-stamped, so you can edit by reading rather than scrubbing. Analysis runs on your individual files, so a synced timeline is never analyzed as a unit. It reflects whatever the files inside it carry. See Edit your video in the transcript and timeline.
Topics and subtopics. Just like the transcript, every file gets its own topics and subtopics. Selects reads the transcript and organizes each file into topics and subtopics, so you can see what the footage covers at a glance. See What are topics and subtopics?.
Speaker detection. Selects identifies who is speaking in each file's transcript. On a synced timeline, matching each speaker to a face and an audio track is a separate step: click [Update] on the [Update data] step in [Sync tools]. Camera switching uses that match later. You can rename, reassign, or add a speaker at any time.
Visual analysis. Selects reads what is on screen and writes a short description for each clip. This is what scene search matches against, so you can find a moment by describing it. See Search your footage: text and scene search.
Shot quality is not part of analysis. Selects can mark the stretches that are shaky or out of focus, per camera angle, but only when you ask chat to check. See Find shaky and blurry shots.
Every file stands on its own. Each video or audio file gets its own transcript, topics, subtopics, and speakers. A synced timeline shows the transcripts of the files inside it, and gets its own topics and speaker profiles when you click [Update] on the [Update data] step in [Sync tools].
When analysis finishes, Selects shows a short notice saying how many clips were analyzed.
Analyze files you add later
When you add files to a project, [Analyze media for edit] opens on its own with the new files checked. Close it, and nothing is analyzed or charged. Files you have analyzed before, in any project, usually come in already analyzed, at no charge.
To analyze a file later, hover it in the project panel and click [Analyze]. The same button appears on any file whose analysis failed, so you can retry just that one. While a file is running you can click [Cancel] on its row to stop it.
You can also analyze from the transcript panel: an unanalyzed file shows [Analyze file], and a draft lists unanalyzed clips with [Analyze]. Selects shows you the files and the estimated credits first, then starts once you confirm. See Add media to an existing project.
Credits
Analysis uses credits from your plan balance:
Video analysis: 1 credit per 10 seconds (6 credits/min, 360 credits/hr)
Audio analysis: 1 credit per minute (60 credits/hr)
You only pay for analysis that succeeds. Files that fail are never charged, files you have already analyzed are not charged again unless you choose to re-analyze them, and stopping partway does not re-charge the files that finished.
Analysis is not the only thing that uses credits. Chatting with the Assistant and generating video, images, and audio use them too. See Edit your video with chat in Selects and Selects app: plans and pricing.
FAQs
Does Selects analyze my footage automatically?
No. Nothing is analyzed until you confirm. When you add files, [Analyze media for edit] opens so you can choose what to analyze. Close it, and nothing runs and no credits are used.
Can I analyze only some of my files?
Yes. Click [Analyze N clips], or hover a clip and click [Analyze], and [Analyze media for edit] opens. From [Analyze N clips], every clip starts checked. From one clip's [Analyze], only that clip starts checked. Check or uncheck clips to choose what gets analyzed. Nothing is analyzed until you confirm.
How long does analysis take?
Roughly 20 minutes for every hour of video, though it varies with your internet connection and your machine. The pill gives you an estimate once it has enough to go on. We are working on making this faster.
Can I use Selects while analysis is running?
Yes. You can move around the app and keep working while it runs. Leave Selects open so it keeps going. If you quit, analysis picks up again when you reopen the app.
What happens if analysis fails on some files?
The rest carry on and finish as normal, and you are not charged for the ones that failed. When the run ends, [Some files couldn't be analyzed] lists them, with [Re-analyze N clips] to try again or [Continue to timeline]. You can also retry a failed file on its own: hover it in the project panel and click [Analyze].
Can I stop analysis and pick it up later?
Yes. Cancel a running file from its row in the [Project] panel. Anything that already finished is kept, and analyzing the rest later does not charge you twice for the finished ones.
What languages does Selects support for transcription?
Selects supports transcription in over 100 languages.
Supported languages: Afrikaans, Albanian, Amharic, Arabic, Armenian, Assamese, Azerbaijani, Bashkir, Basque, Belarusian, Bengali, Bosnian, Breton, Bulgarian, Burmese, Catalan, Chinese, Croatian, Czech, Danish, Dutch, English, Estonian, Faroese, Finnish, Flemish, French, Galician, Georgian, German, Greek, Gujarati, Haitian Creole, Hausa, Hawaiian, Hebrew, Hindi, Hungarian, Icelandic, Indonesian, Italian, Japanese, Javanese, Kannada, Kazakh, Khmer, Korean, Lao, Latin, Latvian, Lingala, Lithuanian, Luxembourgish, Macedonian, Malagasy, Malay, Malayalam, Maltese, Maori, Marathi, Mongolian, Nepali, Norwegian, Norwegian Nynorsk, Occitan, Pashto, Persian, Polish, Portuguese, Punjabi, Romanian, Russian, Sanskrit, Serbian, Shona, Sindhi, Sinhala, Slovak, Slovenian, Somali, Spanish, Spanglish, Sundanese, Swahili, Swedish, Tagalog, Tajik, Tamil, Tatar, Telugu, Thai, Tibetan, Turkish, Turkmen, Ukrainian, Urdu, Uzbek, Vietnamese, Welsh, Yiddish, Yoruba.
Note: Transcription accuracy can vary by language and by audio quality, such as background noise or overlapping speakers.