> For the full page index, fetch https://echonova.ai/llms.txt # Read a transcript Find your way around the Speech to Text Transcript card, play the audio against the text, switch between transcripts, and save a transcript as a file. When a transcription finishes, the **Speech to Text** page shows a **Transcript** card under the upload box: the text, a play button for the audio it came from, and a **Transcripts** list on the right with your latest transcripts. This page walks through that view, so you can read the card, check a passage against the audio, and take the text away as a file. ## What you need - A finished transcription on the page. See [Transcribe a file](/docs/playground/transcribe-a-file). ## Steps ### Step 1: Find your way around The page has two parts: - On the left, the upload box and, under it, the **Transcript** card. The card appears when a transcription finishes and the page scrolls to it. - On the right, **Transcripts**: your latest transcripts, newest first, each with a play button, its title and its length and date. A search box that reads **Search transcripts** narrows the list, and **All in Assets →** opens the **Assets** page on its **Speech to Text** tab. The card shows one transcript at a time: the one that just finished, or the one you clicked in **Transcripts**. ### Step 2: Read the transcript card The top of the card holds: - The title: the file's name, or the first words of the text when the job has no file name. - A line with the length of the file as `m:ss` and the language (**Arabic** or **English**) when it is known. - **Copy**, **Download** and **Clean Noise**, described below. Under the header, a switch with three tabs picks what the card shows: **Text**, **Subtitles (SRT)** and **JSON**. **Text** is open the first time; the card then opens on the tab you used last in this browser. On **Text**, when the transcript was made with **Identify speakers** on, a line under the switch sums up who spoke: each speaker's label, their talk time as `m:ss` and their share of the talking, for example `Speaker 1 1:20 · 62%`. Under it, the text is split into rows, in the order they were said. Each row has a play button, the time it starts at, such as `0:42`, and, when the speaker is known, a chip with the speaker's label: **Speaker 1**, **Speaker 2** and so on, numbered in the order they first speak. Each speaker keeps the same chip style on every row. The labels follow the language of the page, so the Arabic page shows **المتحدث 1**. A transcript made with **Identify speakers** off, or made before speakers were identified, sits in one block with a play button and `0:00` at its start. Arabic text is laid out right to left, and English text left to right. A transcription that came back with no text shows **No speech was found in this file.** in its place. **Subtitles (SRT)** shows the subtitle file itself, the same file **Download** saves from that tab, cue by cue: each cue's number and times, left to right, then its lines in their own direction. While the audio plays, the cue being heard is highlighted. A transcript in one block gives one cue for the whole file. **JSON** shows the JSON file, laid out with one field per line and read left to right. Both scroll inside the card. The first time you open **Subtitles (SRT)** or **JSON** for a transcript, the card shows **Preparing the file…** while the file is built; after that, switching back to the tab is instant. The file is built again when you switch the portal's language, since speaker labels follow it, and for each transcript you open. If the file cannot be prepared, the tab reads **We could not prepare this file. Try again in a moment.**; press **Try again**. ### Step 3: Play the audio against the text The play button on a row plays the audio from the start of that row, and turns into a pause button while it plays; click it again to pause, and once more to carry on from where it stopped. Click a row's time to jump back to the start of that row. While the audio plays, the row being heard is highlighted. On a transcript in one block, the play button at the start of the text plays the audio from the beginning. The play button beside a title in **Transcripts** plays that transcript's audio directly. One clip plays at a time; starting another replaces the first. ### Step 4: Switch between transcripts Click a transcript's title in **Transcripts** to show it on the card. A transcript that is still processing shows **Processing** instead of a date and cannot be opened yet. Transcripts that failed are not listed; they stay visible on the **Assets** page. ### Step 5: Save the transcript Press **Download**. It saves the tab that is open: **Text** as `TXT`, **Subtitles (SRT)** as `SRT` and **JSON** as `JSON`. The file is named after yours with the extension removed and the format's extension added, so `interview.mp3` gives `interview.txt`, `interview.srt` or `interview.json`. The button shows a spinner while the file is prepared. - `TXT` is the plain text. - `SRT` holds subtitles. Each cue lasts at most 7 seconds and holds at most two lines, and starts with its speaker in brackets, such as `[Speaker 1]`, when the speaker is known. Cue times inside a row are estimated from the length of the text, so they are close but not exact. A transcript in one block gives one cue that covers the whole file. - `JSON` holds the id, the file name, the length, the language and the text, then the speakers with their talk time and the timed rows with their speakers. Speaker labels in the file follow the language of the page: `Speaker 1` on the English page, `المتحدث 1` on the Arabic page. **Copy** puts what the open tab shows on the clipboard: the plain transcript on **Text**, the subtitle file on **Subtitles (SRT)** and the JSON on **JSON**. The button reads **Copied** for 2 seconds, or **Could not copy** when the browser refuses. To take a passage rather than the whole text, select it in the card and copy it as you would on any web page. The text on the page cannot be edited; make your corrections in the downloaded file. **Clean Noise** opens **Sound Cleaning** with this transcript's audio already chosen, so you can clean a noisy file and transcribe it again. ### Step 6: Come back to it later Leaving the page clears the card. The **Transcripts** list on the right shows your latest transcripts again the next time you open the page, and **Assets** lists all of them. ## Worked example You transcribe `interview.mp3`, a 9-minute talk between a host and a guest, with **Identify speakers** on, and then press **Record** and record a 45-second note. After each one finishes, the card shows it: first the interview with `9:00` in its header line, a talk-time line such as `Speaker 1 5:40 · 63%` and `Speaker 2 3:20 · 37%`, and rows that start **Speaker 1** at `0:00` and **Speaker 2** at `0:12`; then the note with `0:45`. Both are in **Transcripts**. Click `interview.mp3` there and the card switches back to the interview. Click `0:12` to hear the guest's first answer. Open **Subtitles (SRT)**: the cues start with `[Speaker 1]` or `[Speaker 2]`. Press **Download** and `interview.srt` is saved. ## What happens next - Each further upload or recording adds a transcript to **Transcripts** and shows it on the card. - Leaving the page clears the card. The transcripts and audio remain in **Assets**. ## Troubleshooting ### The card reads "No speech was found in this file." The transcription finished with no text. Check that the recording contains speech at an audible level, then transcribe it again. The job is charged as a completed transcription. ### The downloaded file is not plain text The **Subtitles (SRT)** or **JSON** tab was open, so **Download** saved that format. Open **Text** and press **Download** again. ## FAQ ### Are the speakers in a recording told apart? Yes, when **Identify speakers** is on, which it is by default. Each row carries a speaker chip, and the talk-time line shows how long each speaker spoke. Transcripts made with it off, and older transcripts, stay one block. ### Can I get a timestamp for each sentence? Each row shows the time it starts at, and the **Subtitles (SRT)** tab and its download have a cue every few seconds. The times inside a long row are estimates. A transcript in one block has a single time, `0:00`, and one cue for the whole file. ### Can I edit the transcript on the page? No. Download the text and edit it in your own editor. ### Which formats can I export? `TXT`, `SRT` or `JSON`: open **Text**, **Subtitles (SRT)** or **JSON** and press **Download**. The **Assets** page offers the same three formats. ### Is the transcript kept if I close the tab? Yes, in **Assets**. Only the card on this page is cleared. ## Related - [Transcribe a file](/docs/playground/transcribe-a-file): the steps that lead to this view. - [Speech to Text formats](/docs/playground/stt-formats): the audio formats the page accepts. - [Immediate and queued transcription](/docs/playground/sync-versus-async): what happens to a transcript if you leave before it arrives. - [Speech to Text limits and credits](/docs/playground/stt-limits-and-credits): what each transcript cost.