Can ChatGPT watch videos?

The short answer

Partly. OpenAI's help center says ChatGPT can accept video files as attachments, but it warns the analysis can be incomplete or inaccurate: it may not cover the entire video or interpret the audio accurately. The page doesn't say what ChatGPT reads from the file. For feedback you can rely on, give it stills and a transcript, or connect Codex on your Mac, such as ChatGPT's Work mode, to an editor like Woven Video.

Your options

Attach the video file. OpenAI's image inputs FAQ says video files can be attached through supported upload methods, including on Free plans, and that availability varies by platform, method and account. If you can't pick a video in Photos, try attaching it through Files. Video counts toward your plan's file-upload limits. OpenAI's separate article on uploading files and audio covers documents, spreadsheets, presentations, images and audio; video isn't in its table.

Upload the sound on its own. On a paid ChatGPT plan, audio files up to 512 MB can be transcribed and discussed, in formats such as MP3, M4A, WAV and FLAC. The uploads page says MP4 and WebM files identified as video aren't supported as audio uploads, so save the audio track as its own file first. It also warns that transcripts can contain errors.

Send stills and the words. Image inputs take PNG, JPEG and non-animated GIF files up to 20 MB each. Export a frame from the first second, one from every new shot and the closing frame, then paste the script or a transcript with them. A still can't show a camera move or carry sound, so keep your questions to what one frame or the words can show.

Share live camera video in Voice. OpenAI's Voice page says video and screen sharing are available to eligible subscribers in the iOS and Android apps when you use Advanced, and that the newer Live option doesn't support them. It shares live camera video during a conversation; the page doesn't describe it as a way to review a saved file.

Connect an editor. Woven Video, below. It lets ChatGPT's Work mode ask for stills and a transcript of your footage, then make the edit.

Show ChatGPT your footage through Woven

Woven Video is an AI video editor for Mac. Its connector hands a connected AI app the parts of a video it can read: still frames, the speech as text with a time on every word, and the name of the song. Motion and sound never reach the app, and Woven tells it to admit what it couldn't check.

• ·Through Work mode. After setup, Work mode can pull a video in from a YouTube, TikTok or Instagram link, request up to six frames per call, get a transcript, split off the audio track and look up the song. Sign-in walls, rate limits and region blocks can still stop a download.
• ·Look at the cut, too. Work mode can also request up to six stills of the edited reel. Woven draws them with its own renderer, so pictures, text, captions and effects all show. Export stays blocked until the latest visuals have been rendered.
• ·Or skip ChatGPT for a breakdown. In Woven's chat, paste the link to a video you want to learn from. The built-in agent extracts frames and contact sheets, transcribes the speech, names the song and saves a breakdown of the hook, pacing and shot list, covering the first 90 seconds of anything longer. It then offers to build a reel from it and waits for your yes.
• ·Run that chat on your ChatGPT plan. In Settings, open Models, then Model access, and choose ChatGPT Codex to run Woven's chat on a paid ChatGPT plan. Transcription, captions and generation still use Woven credits.
• ·Credits for transcripts. Transcription is paid from Woven credits. Woven keeps each file's transcript, so transcribing or captioning that file again is free for the seconds already done.

Ask what stills can answer. Good questions: does the first frame make sense with the sound off, is the on-screen text readable at phone size, and does each line of the script sit on the right shot. Leave music and pacing out: no sound or motion arrives, and word timings only show the rhythm of the speech.

Connect ChatGPT's Work mode to Woven

1. 1Pick the app. Go to Woven's Settings, open AI apps and set the AI app menu to Codex (ChatGPT).
2. 2Paste the prompt. Click Copy setup prompt and paste it into Codex on your Mac: Work mode or Codex in the ChatGPT desktop app, the Codex CLI or the Codex IDE extension. The prompt registers Woven's connector with that app, then brings up Woven's Google sign-in.
3. 3Load the tools. When Woven says it is connected, click its Restart button in the ChatGPT desktop app. That quits the app first, so let running tasks finish. In the CLI, start a new session.
4. 4Ask it to look. Point it at a file in your workspace or paste a link, ask what the first frame shows and whether the captions are readable, then ask for the edit.

Requires macOS 15 or later, a signed-in Woven account on a plan or the 3-day free trial, and Codex running on your Mac. The setup prompt stops in ChatGPT's regular chat and in Codex cloud tasks.

FAQ

More answers

Want the editing side, or another assistant's answer? See our answers for Can ChatGPT edit videos?, Can Claude watch videos? and Can Gemini watch videos?.

Sources

• Uploading files and audio to ChatGPT (OpenAI Help Center), checked October 20, 2026
• ChatGPT Image Inputs FAQ (OpenAI Help Center), checked October 20, 2026
• ChatGPT Voice (OpenAI Help Center), checked October 20, 2026