Not according to OpenAI. As of 19 October 2026, OpenAI's upload help page lists documents, spreadsheets, presentations, images and audio, and says MP4 and WebM files identified as video are not supported. Gemini accepts short video uploads, but Google's help page describes answers, summaries and insights about the content, not a video file with captions on it. Claude's own upload page lists no video format. The one route we know of that ends with a captioned MP4 inside a chat is Claude connected to a caption tool, such as the CapsAI connector described below.
So the useful question is what each assistant can do on the way to captions, and where you still need a caption tool. This post takes each in turn, using each vendor's own help pages as read on 19 October 2026. We did not run our own ChatGPT or Gemini tests for this post; the Claude route is one we ran end to end.
First: post text or captions on the video?
"Caption" means two different things. A post caption is the text under a Reel, a TikTok or a YouTube Short: the hook, the hashtags, the call to action. ChatGPT, Gemini and Claude all write that, because it is just text. Paste in your script or describe the clip and ask for three options.
Captions on the video are the words that appear on screen in time with the speech. They need two things an assistant does not do by itself: a transcript with timings, and a step that puts the text on the picture, either burned in (open captions) or as a subtitle file the platform shows on top (closed captions). The rest of this post is about that second kind. Burned-in vs closed captions explains the difference if you are not sure which you need.
The short answer for each assistant
| Assistant | Takes a video file? | What you can get from it | Returns a captioned video? |
|---|---|---|---|
| ChatGPT | No, video is not a supported upload type | A transcript of an audio file on paid plans, which you can ask it to lay out as SRT text | No |
| Gemini app | Yes, up to 2 GB per video; 5 minutes in total on free, 1 hour on Google AI Pro or Ultra | Answers, summaries and insights about the video, per Google's page | No |
| Claude | No, its upload page lists documents and images only | Help with text | Only through a connector to a caption tool |
| Claude with the CapsAI connector | Through an upload link, up to 10 minutes per video | A captioned MP4 in one of 37 styles, plus the original transcript as SRT | Yes |
Each row is sourced in the sections below. These products change often, so check the linked page before you plan around a limit.
ChatGPT: audio in, text out
OpenAI's article on uploading files and audio to ChatGPT lists the supported formats: DOCX, PDF and TXT documents; XLSX, XLS, CSV and TSV spreadsheets; PPTX; images; and audio in WAV, MP3, OGG, FLAC, AAC, M4A and audio-only MP4 or WebM. It then says that "WebM and MP4 files identified as video are not supported by audio uploads." Audio uploads also need a paid plan: the article says they "are not available on the Free plan at this time". Free users are limited to 3 file uploads a day.
So the ChatGPT route to captions looks like this:
- Export the audio from your video as an MP3 or M4A file.
- Upload it on a paid plan and ask for a transcript.
- Ask ChatGPT to lay the transcript out as SRT, with numbered cues and start and end times.
- Check the words and every timestamp against the audio. OpenAI's page warns that transcripts may contain errors and that speaker identification may be unreliable, and it makes no promise about timing.
- Add the SRT to your video in an editor or a platform's caption upload, or burn it in with a free tool, as shown in burning subtitles with HandBrake, VLC or FFmpeg.
That works, but steps 4 and 5 are where the time goes. If you are not sure which subtitle file a platform wants, SRT vs VTT vs TXT covers it.
Gemini: it reads the video, then writes
Gemini is the one assistant here that accepts video. Google's page on uploading and analyzing files in Gemini Apps says each video can be up to 2 GB and that "Total video length can be up to 5 minutes", extended to 1 hour on Google AI Pro or Google AI Ultra. Audio gets 10 minutes, or 3 hours on those plans. The limit is given as total length, not per clip.
The same page describes what Gemini gives back: "answers, summaries, and insights about your content". It does not mention transcripts, subtitle files or timings, and it does not describe returning an edited video. We did not test Gemini for this post. You can ask it what is said in a clip, but treat any text or SRT it writes as an unchecked draft. You then have the same two jobs as with ChatGPT: check every word and timing, and get the captions onto the video with another tool.
Claude: not on its own, but with a connector
Anthropic's page on uploading files to Claude lists document types such as PDF, DOCX, CSV and TXT, and images in JPEG, PNG, GIF and WebP. There is no video or audio format on the list, so the page offers no way for Claude on its own to transcribe or caption a video file.
What Claude does have is connectors: links to outside services that Claude can use from a chat. Anthropic's custom connectors help page says custom connectors work on the Free, Pro, Max, Team and Enterprise plans, and that Free users are limited to one custom connector. On Team and Enterprise, an owner adds the connector to the organization first.
Claude with the CapsAI connector
CapsAI has a custom connector for Claude. You add it once, under Customize, then Connectors, then Add custom connector, and sign in with your CapsAI email. The CapsAI for Claude page shows the five setup steps with screenshots. CapsAI is not in Claude's connector directory, which is why it is added as a custom connector.
In a chat, you ask Claude to caption a video. Claude starts a job and gives you an upload link; you open it and drop in the video, which goes straight to CapsAI rather than into the chat. CapsAI transcribes the speech and renders the captions in the style you picked, and Claude gives you a download link for the MP4. What to know before you try it:
- Videos can be up to 10 minutes long, as MP4, MOV, M4V or WebM. Links to YouTube or Instagram videos are not supported, so upload the file.
- Transcription uses your CapsAI minutes, one per minute of video. There are 3 free minutes to start, then one-time packs from $4 (₹49 in India, €4 in the euro area).
- You can pick any of the 37 caption styles from previews in the chat, or ask for plain static subtitles at the top, middle or bottom. The style gallery shows them all; Pro styles need a paid pack.
- The connector can also return the original transcript as SRT. It is the unedited transcript, so read it before you publish it.
- The spoken languages are the 11 CapsAI supports: English, Hinglish, Hindi, Bengali, Telugu, Marathi, Tamil, Urdu, Gujarati, Kannada and Punjabi. There is no translation.
We ran this route end to end from Claude Code on 11 October 2026 and wrote up every command in how to caption a video from Claude Code.
Hinglish and Indian-language videos
OpenAI's page says transcription performance "may vary across languages" without listing which languages work well, and Google's page lists no languages at all. If your video mixes Hindi and English, tell the assistant which script you want, Roman letters or Devanagari, and check every line. CapsAI treats Hinglish as its own language option and writes it in Roman script. The Hinglish caption generator does the same thing in the browser, without Claude.
Which route to use
- You need post text, hashtags or hooks: any of the three assistants.
- You need a transcript or an SRT draft and will fix the timings yourself: ChatGPT on a paid plan, from the audio. Gemini takes the video itself, within 5 minutes in total on the free plan, but Google's page does not describe transcripts, so try it on a short clip first.
- You need styled captions on the video without leaving the chat: Claude with the CapsAI connector.
- You do not use an assistant at all: upload to a caption generator directly; the auto caption generator does the same job in a browser.
What to do next
If a captioned MP4 is the goal, set up the connector from the CapsAI for Claude page, try a short clip on the 3 free minutes, and read the draft before you caption anything longer. If you only need words to paste under a post, any assistant will do, and you do not need a caption tool at all.
