Generating Subtitles

Plenty of videos come with no subtitles at all — old uploads, small channels, livestream recordings. Fetchii can create them from the video's audio, right on your Mac, in whatever language the video speaks. No cloud service, no per-minute fees.

Before You Start

Subtitle generation needs the Speech model, listed under Settings → Intelligence → On-Device Models. Two sizes are available in the row's menu:

If the model isn't downloaded yet, Fetchii fetches it the first time you generate subtitles. Like all Intelligence features, this requires an Apple Silicon Mac. See Getting Started with Intelligence for the basics.

Generating Subtitles for One Video

1. Hover over a completed download and click its sparkles button (✦) — or right-click it and choose Generate subtitles.
2. In the popover, click Generate subtitles.
3. Watch the progress right in the popover — the row's chip reads Transcribing… meanwhile, and you can cancel anytime.

When it finishes, Fetchii saves a standard .srt file next to the video, so the subtitles work in any player. The row gains a Subtitles chip; its popover has a Show Subtitle File button, and the transcript becomes available to search, summaries, and translation.

If the result isn't right — say the wrong language was detected — open the popover's … menu, choose Regenerate Subtitles…, and confirm with Regenerate subtitles. This replaces the current transcript and any translation.

Generating Subtitles Automatically

Prefer a zero-click library? In Settings → Intelligence → Subtitles, turn on Generate subtitles automatically on this Mac. Every new download that arrives without subtitles gets transcribed as soon as it completes.

Videos that already have official subtitles are skipped — Fetchii won't overwrite what the platform provided. (For downloading official subtitles, see Quality & Format Options.) The Intelligence features only work with subtitles Fetchii generated, so if you want to search, summarize, or translate one of those videos, click Generate subtitles for it yourself.

Choosing the Spoken Language

By default Fetchii detects the language it hears automatically. If a video mixes languages or detection guesses wrong, set the spoken language explicitly:

• Globally: Settings → Intelligence → Subtitles → Spoken language
• The picker offers Detect automatically plus the 10 most common languages directly; automatic detection covers the rest of what the multilingual speech model understands.

Why the Results Stay Clean

Speech models have a known weakness: over music or silence, they can "hallucinate" — inventing captions for words nobody said. Fetchii runs a voice-activity pre-filter ahead of transcription, so only actual speech reaches the speech model. Music intros, ambient noise, and quiet stretches simply produce no captions, the way it should be.

Troubleshooting

• Generation sits at "Downloading model" — the Speech model is still fetching. You can watch progress in Settings → Intelligence.
• It sits at "Preparing…" — the first time a model is used, Fetchii prepares it on your Mac, which can take a few minutes. Later runs are much faster.
• Wrong language in the output — set the spoken language manually (see above) and regenerate.
• The subtitles miss quiet dialogue — switch the Speech model to Best quality; it's noticeably better with difficult audio.
• Nothing happens on an Intel Mac — subtitle generation requires Apple Silicon.

Tip: Generated subtitles are the foundation for everything else — once a video has a transcript, it becomes searchable by meaning, summarizable, and translatable.