Support video subtitles. Make the next step clear.

Give one customer a reply they can follow

A personal support video can show the next step for a particular customer’s question. Its captions need to preserve which situation the advice applies to, what to try and when to stop. Polyglot can transcribe an authorized recording locally on an Apple Silicon Mac, help you review and translate the text, and export separate SRT or WebVTT files. Your support system handles delivery and the conversation that follows.

This workflow starts with an approved recorded reply. It differs from an incoming bug reproduction, which documents a failure, and a reusable tutorial, which teaches a general task. A demonstration that works in your test account does not prove the customer’s issue is fixed. Keep that distinction in the spoken reply, captions and accompanying message.

Loom’s support page describes personalized recorded replies as an established support workflow. If your approved recording platform already provides suitable captions and review, that may be sufficient. Polyglot is useful when you have a permitted local recording, need reviewed subtitle files or want to prepare a translated track. It does not record your screen, diagnose a ticket or connect to your help desk.

Confirm the scope before transcribing

Write a short reply note outside Polyglot. Record the question being answered, the relevant product version, the starting condition, the approved action, any stop condition and what the customer should report afterward. Use your organization’s approved support guidance. If the instructions themselves are uncertain, resolve that before making a polished video.

Use recordings and software demonstrations you are authorized to process and share. Prefer an approved demo account with synthetic data. Check the screen, notifications, browser tabs and spoken audio for customer names, private URLs, account details and secrets. Permission to investigate a ticket does not automatically permit publishing its contents as a tutorial.

If excluded information appears, prepare a clean recording or resolve it in suitable editing software before transcription. Deleting a name from captions leaves it in the audio or picture. Polyglot does not redact recordings, manage consent or control access to the files you later share. Keep the reply and exported text in approved locations.

Separate a proposed step from a confirmed result

The fictional example below concerns a preview setting in a made-up document app. It is a caption-review exercise, not a tested product instruction. The agent has demonstrated the step in a demo account; the customer’s outcome is still unknown.

In the recording, the agent says to change a setting only if a particular starting condition holds. The draft loses that condition. Correct the caption to match the audio, then make sure the companion note tells the customer what to do when their screen differs. Do not silently turn a suggestion into an unconditional instruction.

If the spoken advice is wrong or incomplete, revise the recording with the support owner. Accurate captions cannot repair inaccurate advice. Label any additional written clarification as a support note rather than inserting words the agent never said into the timed transcript.

Fictional preview reply
Clip: preview-reply-v02.mp4
Demo app: Sample Pages 4.2

Heard:
“If Preview is selected,
choose Final and try again.”

Draft:
“Choose Final and try again.”
Correction: restore condition.

Support note:
Different screen?
Stop and reply.
Demo outcome: export opens.
Customer: not yet confirmed.
Next check: does yours open?

Review the source and each translated track

Import the final approved recording and confirm that its spoken language is supported for transcription. Listen through the generated text alongside playback. Check product names, menu labels, error codes, numbers and words that limit the instruction: if, only, before, after and not. W3C’s caption guidance treats automatic text as a starting point that needs accuracy review.

A cursor movement or a silent dialog cannot be recovered from speech recognition. If the narration says “select this”, add a clearly labeled written step identifying the control, or record a clearer explanation. Keep meaningful notification sounds and verified speaker labels where needed; Polyglot does not automatically identify speakers, read interface text or describe screen actions.

Translate after the source text is reviewed. Ask a competent language reviewer to check the condition, action, stop point and requested response as a connected sequence. Preserve uncertainty such as “this may help” and distinguish a demonstrated outcome from a confirmed customer result. Translation targets and speech-recognition languages are separate capabilities.

Keep translated instructions connected to the interface the customer actually sees. A translated menu name may not match an English-language screen. Put a short label mapping in the companion note when useful. Timing and cue-structure validation do not establish that a translation is technically correct.

Deliver the exact reply and verify its playback

Export the subtitle format accepted by the approved receiving player. Pair it with the exact recording revision and identify the language clearly. Upload through your normal support workflow; Polyglot does not send the reply or attach files to a ticket. An attachment that uploads successfully is not proof that the customer’s video player can display it.

Open the delivered link using an approved recipient-view test. Confirm the intended access restrictions, the correct video revision and a visible caption control. Enable the track, seek to the conditional instruction and watch through the outcome. Check that text does not cover the control the customer needs to find, including on a narrow screen.

If the destination cannot load SRT or WebVTT, use an approved player that can, or create a captioned video separately in suitable editing software. Polyglot exports separate subtitle files and does not burn them into the picture. Do not move a private support reply onto a public service just to gain a caption control.

Export plain text when a readable companion is useful. Format it in the support reply and add necessary visual context manually, clearly separating it from the spoken words. W3C’s transcript guidance explains that a transcript derived from captions may need additional visual information and should be easy to find. A raw speech transcript may not be enough to follow a screen demonstration.

Finish the reply note with the delivered video and caption revisions, the reviewer and the requested customer response. Track customer confirmation in the support system. A viewed video, an accurate subtitle file or a successful demo is not a resolution receipt. If the instructions change, review the new recording and tracks together before replacing the delivered version.

Keep exploring.