ScribeToAny

Transcribe local audio or video files (MP3, M4A, WAV, MP4, MOV, WEBM…) or public X, Twitch, Loom, Google Drive and Dropbox links into text, SRT or VTT subtitles with the ScribeToAny CLI, with timestamps and optional speaker labels. Use when the user asks to transcribe a recording, meeting, interview, podcast, lecture or video on their computer, turn speech into text, or make subtitles or captions for a media file; also for "转写", "转文字", "做字幕", "语音转文字". Not for editing existing subtitle files.

Transcribe with ScribeToAny

The scribetoany CLI uploads a file to the user's ScribeToAny account, waits for the transcript and saves it next to where you run it. Results go to stdout, progress to stderr. Every command takes --json.

1. Make sure the CLI runs

bash

scribetoany --version

If it isn't installed, use npx -y scribetoany in place of scribetoany everywhere below (Node.js 20+). Ask the user before running npm i -g scribetoany.

2. Make sure the user is signed in

bash

scribetoany auth status --json

Exit code 3 means not signed in. Run scribetoany auth login: it opens the browser, where the user signs in and approves "ScribeToAny CLI", and waits up to 5 minutes. Tell the user to approve it in the browser. If your shell times out sooner, run it in the background and check auth status after. On a remote machine or a sandbox without a browser, sign-in doesn't work yet: ask the user to run the transcription on their own computer, or to upload the file at https://scribetoany.com/dashboard.

3. Transcribe

bash

scribetoany transcribe interview.m4a --format srt --json

• --format txt|srt|vtt|json: txt (default) has [m:ss] timestamps, --no-timestamps drops them. json is the segment list with start/end milliseconds and speakers.
• --language <code> when the user says which language is spoken (Whisper codes: en, zh, ja, es…). Leave it out to detect it.
• --speakers for interviews, meetings and podcasts with several voices; --speakers 3 if the user knows how many.
• --translate-to-english to get English text from non-English speech.
• --out <path> picks the file (or a directory); without it the result is ./<file name>.<format>. An existing file is not overwritten unless you add --force. --out - prints the transcript instead.
• Several files at once: scribetoany transcribe a.mp3 b.mp3 --out transcripts/.

Long recordings take several minutes after the upload. For anything over about 10 minutes, don't block your shell:

bash

scribetoany transcribe lecture.mp4 --no-wait --json   # prints the id
scribetoany get <id> --wait --format srt --json       # run in the background

With --json, transcribe and get print { id, name, status, language, durationSec, planLimit, url, output }, where output is the saved file's path. planLimit is set when the recording was longer than the plan allows per file and only its start was transcribed: always tell the user, using its message.

4. Use the result

Read the saved file; for long transcripts, read it in parts instead of printing all of it into the conversation. url opens the transcript in the ScribeToAny editor, where the user can fix words and speakers.

Other commands: scribetoany list --json (recent transcripts and ids), scribetoany get <id> --format vtt (any finished transcript again, in any format), scribetoany usage --json (plan, files today, minutes left).

Exit codes

Code	Meaning	What to do
0	done
2	bad input (missing file, unsupported type, output exists, bad option)	fix the command
3	not signed in or sign-in expired	scribetoany auth login
4	plan limit: daily files, monthly minutes or too many at once	show scribetoany usage; plans at https://scribetoany.com/pricing
5	transcription failed	show the error to the user; don't retry blindly
6	network or server error	retry once, then report
7	not finished yet	scribetoany get <id> --wait

Rules

• The file is uploaded to ScribeToAny and counts toward the user's plan. Check scribetoany usage before a large batch. Ask first when the user hasn't asked for a transcription, or the recording looks confidential.
• YouTube, TikTok and Instagram links can't be fetched: download the video first (with the user's say-so), then transcribe the file.
• Never print, log or copy the token, SCRIBETOANY_TOKEN or the credentials file (~/.config/scribetoany/credentials.json).