Harmar: AI Subtitles

Description

From the marketplace listing

Harmar adds subtitles to your videos. Attach a video and Harmar transcribes the speech, times every word, and returns a captioned MP4 for Instagram Reels, TikTok or YouTube, or an SRT/VTT subtitle file for your video editor. Made for speech that mixes languages. Many people switch languages on camera: Hinglish, Spanglish, Taglish, French or Portuguese with English, Armenian, Ukrainian or Kazakh with Russian. Harmar transcribes mixed speech as it was said, writing each word in the language and alphabet it was spoken in. On real code-switched recordings it measured a 6.6% average word error rate. Works in 63 languages, with the spoken language detected automatically. On the public FLEURS benchmark it measured an 8.1% average word error rate across 57 languages, including languages such as Georgian, Armenian, Uzbek, Kazakh, Kyrgyz, Tajik, Mongolian, Maltese, Swahili and Yoruba. Per-language results and method: https://harmar.ai/en/blog/whisper-accuracy-benchmark-59-languages Useful for client videos, interviews, podcasts, courses, ads and short-form content. What you can do: • Get subtitles for a video you attach or share as a link. • Translate the subtitles into another supported language, giving the video a second subtitle track. • Choose a caption style (karaoke word highlights, pill highlights, pop-in words or classic subtitles), adjust colours and fonts, and save it for later videos. • Download the captioned MP4 or an SRT/VTT file. • Open any video in the Harmar web editor to change a word or adjust timing. Harmar works with your Harmar account; you sign in the first time you use it. Harmar does not generate video, audio or images: it transcribes the speech in your own video and renders it as subtitles.