Commentary over a loud game

Speech recognition works on the mixed audio of your video. In a gameplay recording that mix holds your commentary, teammates on voice chat, the soundtrack, gunfire, engines and in-game characters talking. AI subtitles try to transcribe every voice they hear, so you get your own words plus cutscene dialogue plus a teammate's callout in one stream, with errors wherever effects bury the speech.

The cleanest fix happens before you open the editor. When you export from your editing app, bring the commentary forward and duck the game under it, and export the clip with a single mixed audio track where your voice sits clearly on top. Recording software such as OBS can capture the microphone and the game on separate tracks, which makes that balance easy to adjust later. Speech recognition and background noise explains why loud effects cause dropped and invented words.

After the AI pass, decide what stays. Many games already subtitle their own dialogue, so delete those lines to avoid doubling up. Keep a teammate's line only when it matters to the moment.

Finding space the HUD doesn't use

Every game puts its interface somewhere different, and bottom centre, the usual subtitle spot, is exactly where many games place health bars, ammo counts, cooldowns and their own subtitles. Spend a minute looking at your footage before styling:

Shooters: Bottom centre and corners are busy with ammo, health and the minimap, and kill feeds often sit top right. Top centre, or just above the middle, tends to stay clear.
MOBAs and strategy: The bottom third is a dense control bar. Top, or middle with a small upward offset, is usually safer.
Racing: Speedometer and map take the bottom corners, so a centred bottom position raised slightly with the offset often clears them.
Open-world and story games: Clean screens while exploring, but the game's own subtitles appear at the bottom during dialogue. A top position avoids stacking two subtitle tracks.
Facecam layouts: If your webcam sits in a corner, centre alignment and a sensible max line width keep text off your face.

Position is a single setting for the whole video, top, middle or bottom plus a vertical offset, so choose the area that clears the HUD for most of the runtime. Moving subtitles to the top goes further for footage where the bottom isn't an option.

Gamertags, callouts and slang

Speech recognition learns from general speech, and gaming vocabulary is anything but general. Expect it to stumble on:

Gamertags and streamer names, which turn into ordinary words or vanish.
Map callouts, weapon, agent and champion names, and ability abbreviations.
Slang, memes and the in-jokes your community uses.
Shouting and laughing over another voice, which garbles both.

Correct names as you find them, and keep a short list for each game you play so the next review goes faster. Swearing is your call: leave it, mask it with asterisks, or cut it, but stay consistent through the video, particularly if the channel is monetised. Fixing auto-generated subtitles has a review order that catches the worst mistakes first.

Subtitles that keep up with fast play

Commentary in a tense moment comes fast. The editor warns on any line faster than 21 characters per second and any line on screen for less than 0.8 seconds, and in gaming clips those warnings bunch up around clutch plays and jump scares. There are three fixes: trim words without changing what you meant, merge two short lines into one that stays up longer, or stretch a line's end a little past the speech when the next line doesn't follow straight away.

The Creator preset this editor starts with, bold yellow text inside a thick outline, suits busy, high-contrast footage. If yellow clashes with your channel colours, keep the heavy outline and change the text colour; the outline is what keeps words legible over explosions and particle effects. UPPERCASE adds energy but slows reading on long lines, so save it for short bursts of commentary rather than long explanations. Yellow subtitles explains where that colour choice comes from.

Long sessions and high resolutions

AI subtitles accept videos up to 2 hours, so a stream VOD longer than that needs cutting into parts first, and in practice the highlights are usually what deserves subtitles anyway. A 20-minute video costs 20 credits (2¢), and signed-in accounts get 10 free credits every day.

Rendering keeps the source resolution and frame rate, so 1440p60 or 4K60 gameplay comes back at 1440p60 or 4K60 with no downscaling. Bigger files take longer to encode on your own machine and need free disk space. If you captured in HDR, the output is converted to SDR and the editor warns you first; colours can look flatter than in the HDR original. Subtitling 4K video covers what to expect from large renders.

Three typical uploads

12-minute ranked match, 1080p60, commentary only

Duck the game audio in the export, run AI subtitles, correct agent names and callouts, keep Creator at the top centre clear of the ammo and health display, and render.

40-minute relaxed building let's-play

The game audio is quiet, so the commentary transcribes cleanly. Switch to Minimal or Modern for a calmer look and raise the subtitles with the vertical offset so they sit above the hotbar.

Vertical highlight from a stream for Shorts

Crop to 9:16 in your editing app first, then subtitle. Turn on the 9:16 safe-area guide and keep text away from the bottom and right edges; YouTube Shorts subtitles covers that layout.

Reaching players in other languages

Gaming audiences are spread across the world. AI subtitles can translate your commentary into one of 21 languages, a practical way to give a clip a Spanish, Portuguese or Japanese subtitle version. Slang and memes rarely survive translation intact, so have a fluent viewer check anything that rests on a joke. Translating gaming videos covers community translation, dubbing and titles. If all you need is a subtitle file to upload to YouTube rather than a rendered video, the subtitle generator produces one directly.

About the subtitle editor on this page

Opens: MP4, MOV, M4V, WebM, MKV videos your browser can decode
Subtitles from: AI (from the speech, optionally translated into one of 21 languages), an SRT or VTT file, or typing them in
You download: An MP4 with the subtitles drawn into the picture at the original resolution and frame rate, plus SRT and VTT files
Cost: Editing, styling, rendering and SRT/VTT export are free, with no account and no watermark. AI subtitles cost 1 credit per minute (5-credit minimum per video); signing in with Google gives 100 free credits and accounts get 10 free credits a day
Privacy: The video never leaves your device and is rendered in your browser. Only for AI subtitles is the audio sent, over HTTPS, and it is deleted within 30 minutes of delivery
Rendering needs: Chrome or Edge 94+, Safari 16.4+, or Firefox 130+ on a computer. Editing and SRT/VTT export work in any current browser

More: subtitle and caption generators

Subtitle tools

Frequently asked questions

Why do game sounds cause errors in my subtitles?

The AI transcribes the full audio mix, so loud effects and music can mask your words, and in-game voices get transcribed too. Lower the game audio under your commentary before exporting the clip and the draft gets much cleaner.

Can the subtitles move to dodge the HUD in different scenes?

No. Position is set once for the whole video, so choose the area that stays clear for most of the clip. Top centre is often the safest choice for shooters and strategy games.

Will the rendered gameplay keep 60 fps?

Yes. The output keeps the source frame rate and resolution, including 60 fps and higher-resolution captures. It is a re-encode at a generous bitrate, so check fast motion in the result before uploading.

Can I subtitle a full stream VOD?

AI subtitles accept up to 2 hours per video, so a longer VOD has to be split first. Many creators subtitle only the highlight clips, which is quicker to review and costs fewer credits.

Will my teammates on voice chat be subtitled?

If their voices are in the audio, the AI will try to transcribe them. Keep their lines when they matter or delete them; there are no speaker names, so add a name prefix if viewers need to know who is talking.