How to transcribe a Twitter or X video
Open the post with the video attached and copy its status URL. Older twitter.com links and current x.com links identify the same kind of input. A profile, a search page or the surrounding thread is not a substitute for an individual video post.
Identify the exact recording
Play the video to confirm it is the clip you need. If a post contains several videos, note that transcript1 processes the first video only. Choose an original file upload when you need a different recording you have locally.
Submit the post link
Paste the URL into the transcript1 dashboard and start a transcription. Automatic language detection is the default. Request speaker identification if separating voices will help you review an interview or exchange.
Check the important passage
Open the completed result, locate the relevant line and replay that part of the video. Listen to the preceding question as well as the answer; qualification and negation can change the meaning of a short quote.
Save text with its source
Copy the reviewed passage or download the transcript. In your notes, keep the post link, timestamp and any correction you made after listening. The transcript alone is not a complete source record.
Which X posts can be transcribed?
The supported input is a public video post with a URL such as x.com/user/status/1234567890 or the equivalent twitter.com address. That example illustrates the URL structure, not a recording to test. The video must be accessible and have an audio track. A text-only post has no spoken content to extract.
Protected or otherwise login-required content cannot be fetched. Profiles, live pages and stories are not supported. This is a video-post workflow, not an integration for joining or recording Spaces. If you have an audio recording in a supported file format, you can submit the file separately.
Only the first video in a multi-video post is processed. A transcript of that video should not be described as a transcript of the whole thread or every attached clip. Verify the selected media before using a result in a comparison or quotation archive.
Use timestamps to check a quote, not just find it
A transcript is a useful search aid, especially when a longer interview is shared as a clip. It is also a machine-generated account of the audio. Before quoting a sentence, listen for proper names, numbers, hedges and short words such as “not.” Check that the speaker label matches the voice at that point.
Preserve enough surrounding language to make the quotation understandable. If the clip begins halfway through a sentence, the transcript cannot recover the missing start. If the post contains an edited excerpt, transcription cannot establish what happened outside that excerpt. Note those limits in the working record.
For an illustrative review note, record “statement at 00:18; question begins at 00:12; name checked against audio,” followed by the source URL. This is your editorial annotation, not an automatically generated verdict about the statement. transcript1 provides words and timing; it does not verify whether a speaker’s claims are true.
Post text, subtitles and speech may disagree
A post’s written framing may paraphrase the video or emphasize only one part of it. Burned-in subtitles may be edited, translated or incomplete. transcript1 transcribes the soundtrack with AI speech-to-text rather than importing either of those text sources. Keep the post’s description separately if it matters to your analysis.
When comparing a transcript with subtitles visible in the clip, inspect the audio before deciding which is right. Compressed audio, multiple voices and background sound can all introduce recognition errors. A clear-looking subtitle is not proof of the wording, and a fluent-looking transcript is not proof either.
Text displayed on charts, screenshots or presentation slides is outside the audio transcript. Include visual evidence in separate notes with its own location in the recording. This keeps spoken quotations distinct from what the video shows.
Choose an export for notes, an archive or a clip
TXT is the simplest format for a reading document. JSON retains segments and start/end seconds for a searchable application. Store the original post URL and the transcription id in your own record so you can revisit the result without submitting another job.
SRT and WebVTT are useful when preparing captions for a video you can edit and publish. Import the file into your editor or player, then review the timing and text. If your clip starts later than the submitted video, the original times need to be shifted or regenerated for the final edit.
For research across several posts, create a separate transcription for each video and label every result clearly. Avoid concatenating transcripts into an unlabeled document: identical speakers or repeated clips can otherwise be mistaken for independent statements.
Automate video-post transcription
Send the status URL to the REST API; platform is detected automatically, or you can pass x explicitly. The API uses speech-to-text for X videos. Keep the returned job id and read that job while it processes. A completed job can be fetched in several formats without using additional minutes.
An MCP-connected assistant can retrieve the same transcript for analysis. Instruct it to cite timestamps, read the whole relevant passage and distinguish quotations from paraphrases. This is particularly important when the input is a short clip whose wider context is not available in the transcript.
Successful transcription costs one credit per media minute, rounded up. A verified new account receives 60 free minutes once. Failed and canceled jobs are not charged, including failures caused by unavailable media or login requirements.