Skip to main content

Grab the text in a screenshot. Record with sound.

Text in a foreign app, a game menu or a chat screenshot is just pixels you can’t copy, and a recording with no sound, or with only your mic, is a wasted take. 77Fox for Windows grabs text with Alt+A and puts translations back where the original was, records your screen with Ctrl+Alt+R while giving system audio and mic separate levels, and converts files like NCM, HEIC and WEBP on your own PC.

77Fox’s screen capture is a set of Windows desktop tools: Alt+A screenshots with annotation, on-device text recognition and translations placed over the original text, Ctrl+Alt+R screen recording with separate system audio and microphone levels, and conversion for images, videos and NetEase NCM files.

Download for Windowsv6.9.2 · 426 MB · Windows 10 / 11 · 64-bitWhat this page covers

Alt+A screenshots

Press Alt+A and the screen freezes, so you can take your time drawing the area you want. Hover over a window or a button and the selection snaps to its edges; scroll to switch between the outer window and the controls inside it. Move or resize the selection as you like, on any of your monitors.

Press Enter or Ctrl+C to copy the shot to the clipboard, or Ctrl+S to save it as a PNG named after the time you took it. Right-click or Esc cancels the whole screenshot.

Screenshots and recording are on by default, and the shortcuts can be changed under Settings → Tools. If another app already uses a shortcut, 77Fox tells you about the conflict instead of taking it over. Screenshots include the 77Fox window by default; turn on “Hide 77Fox during screenshots” to leave it out.

Annotation

Mark up a shot before you send it: pen, arrow, rectangle and ellipse, plus a select tool for picking an annotation you’ve already drawn. Hold Shift for a square or a circle.

Color and stroke width live in their own small floating panel, shown only while a drawing tool is active: six colors (red, yellow, green, blue, white and black) and a stroke slider from 1 to 20. Changes apply to the annotations you draw next, and your choices stay put when you switch tools.

Ctrl+Z undoes, Ctrl+Shift+Z redoes, and Delete removes the selected annotation. When you copy or save, the annotations are merged into the image.

Extract text

Click Extract text in the screenshot toolbar and 77Fox reads the text on your own PC, then shows it in a box you can copy from. You can also drag across a few lines right on the screenshot and press Ctrl+C to copy just those. Lines that only wrap are joined into paragraphs; lists and tables are split row by row at separator lines and bullet markers.

The first time, it downloads a recognition model of about 31 MB. After that, recognition works offline. The screenshot itself is never uploaded.

Chinese and English recognition has been tested; other languages depend on what the model can read, so there’s no promise for them. Vertical text, clearly tilted text and the reading order of complex tables aren’t handled yet, and an oversized selection asks you to pick a smaller area.

Screenshot translation

Click Translate and 77Fox recognizes the text, then lays each translated paragraph over the original: a block in a color sampled from the image covers the old text and the translation is written on top. It’s a flat overlay, not photo retouching, and it doesn’t recreate the original font. When a translation doesn’t fit, Full translation opens beside the selection with every paragraph and its original, ready to copy in one go. Switch between original and translation anytime; copied and saved images include the translation and your annotations.

The translation bar sits just outside the screenshot toolbar. Engines come in five groups: a free channel (Google’s public endpoint), 77Fox Cloud, your own API key (DeepL or OpenRouter, for example), local models, and command-line AI tools on your PC such as Codex or Claude Code (slower). A source that isn’t set up yet links straight to the setting it needs. The source, model and target language you pick, plus your three most recent choices, become the defaults for the next screenshot. If you hit a rate limit you see a countdown and can switch engines; stopping keeps the paragraphs already translated.

There are nine target languages: Chinese, English, Japanese, Korean, German, French, Italian, Spanish and Hebrew. The recognition limits apply here too, so vertical text, tilted text and complex tables won’t translate well. Text recognition is free and so is the free channel; 77Fox Cloud needs a paid plan and uses AI credits (from £1.49/month, see pricing); your own key, local models and local CLI tools use no 77Fox credits.

Screenshot pixels never leave your PC. When you click Translate, only the recognized text goes to the engine you chose, and with a local translation model the text stays on your PC too.

Screen recording with audio

Press Ctrl+Alt+R to open the recorder and pick the full screen, a window or a region. Recording starts only when you click the record button; the shortcut just brings up the panel. Ctrl+Alt+P pauses and resumes, and paused time is left out of the file. You get an MP4, saved to Videos / 77Fox by default and named after the start time, and an existing file is never overwritten.

Before your first recording, install the encoder: when the recorder asks, click Install encoder and confirm the ffmpeg download (about 106 MB) in Settings. You only do this once; after that, recording works offline, and video conversion uses the same copy.

Sound comes from two sources: system audio is on by default and the microphone is off. Each has its own 0–200% level, which changes only the recording, not your system volume. With the settings open you see live waveforms for both, so a source that isn’t picking anything up is obvious before you start. The screen list shows each display’s resolution and refresh rate.

Choose H.264, H.265 or AV1, encoded on the GPU or the CPU. The defaults are original size, 50fps, 12 Mbps, H.264 on the GPU. Resolution can be the original size, 720p to 2160p, or custom; frame rate comes from presets between 15 and 200, plus any values you add; bitrate runs from 1 to 200 Mbps. Being able to pick a value doesn’t mean your PC can record it: high resolutions and frame rates depend on your graphics card and the encoding path. On our test machine, GPU encoding at 1440p and 4K allowed at most 30fps, while 1080p went up to 120fps. For high frame rates, start at 1080p and record a 10-second test first. Combinations your hardware can’t handle are marked Not supported (for a window, the check runs when you click record), and settings are never lowered behind your back.

If the app crashes or the power goes out mid-recording, “Recover interrupted recording” brings back the parts already written to disk; the last moments that weren’t written yet may be lost. Get consent before you record other people’s voices, calls or meetings.

Recording games

When you open the recorder, 77Fox looks for running games. If there’s exactly one, its window is preselected; if there are several, the one in front when you pressed the shortcut wins; if it can’t tell, the choice is left to you. Detection relies on Steam and Epic install records and the window traits of Unity and Unreal games, not window titles, and leaves out launchers and editors. Other apps’ windows are never preselected for you.

Once a game recording starts, the recorder fades out after 3 seconds without interaction so it stays out of the shot, and a red dot on the tray icon shows you’re recording (gray while paused). Press Ctrl+Alt+R or use the tray to bring the recorder back; it won’t steal focus from the game.

Games are captured through Windows window capture, with nothing injected into the game process. Recording covers only what happens after you press record, so anything before that isn’t kept. The high-frame-rate advice above still applies: start at 1080p and check a 10-second test.

Format conversion and NCM

The Image & video converter is off by default; turn it on under Settings → Features. Drop files into the converter (images and videos can be mixed) or onto its dashboard widget to convert a whole batch. You can also add it to the Windows context menu and right-click files in File Explorer to pick items like “To PNG” or “To MP4”. That menu is off by default too: switch each format on or off so only the ones you use appear. The labels are always in English, and the entries are written only for the current Windows user, so normal user rights are enough. Converted files land next to the originals, with a number added if the name is taken, so nothing gets overwritten.

Images and video

Images: PNG, JPG/JPEG, WEBP, AVIF, GIF, BMP, TIF/TIFF and HEIC in; PNG, JPG, WEBP or AVIF out, with quality and longest-edge options. iPhone HEIC photos need ffmpeg; every other image format converts with nothing extra to download. Video: 14 input formats including MP4, MKV, MOV, WEBM and AVI; out to MP4, WEBM, MKV, MOV or GIF, or pull the audio out as MP3 or WAV. Pick a resolution from the original down to 360p and one of three quality levels: High quality, Balanced or Smaller file. Video and HEIC conversion use ffmpeg, which downloads the first time only after you confirm (about 106 MB, a fixed version that’s checked before use). It stays on your PC, isn’t synced, and works offline from then on; screen recording uses the same copy, so one install covers both.

NetEase NCM

“Original (MP3/FLAC)” takes the MP3 or FLAC stored inside an .ncm file exactly as it is, with no re-encoding and no ffmpeg, so the audio is identical to the original. If the file carries no tags, the title, artist, album and cover are written in; existing tags are left as they are. You can also convert to MP3 or WAV; when that needs re-encoding, ffmpeg is used.

Use this only on files you have the right to use, for personal use only, and don’t share the converted files; respect copyright and the platform’s terms. Everything is processed on your PC, nothing is uploaded, and it’s free.

Blog

Translate a screenshot in place

Capture part of the screen with 77Fox and put the translation back where the original text was.

  1. Press Alt+A — Drag out the area you want, or let the selection snap to a window.
  2. Click Translate — Click Translate in the screenshot toolbar. The first time, you’re asked to download the recognition model (about 31 MB).
  3. Pick an engine and language — Choose an engine and target language in the translation bar. By default it uses the free channel (Google’s public endpoint) and your interface language.
  4. Copy or save — Press Enter or Ctrl+C to copy the translated image, or Ctrl+S to save a PNG. For text only, select it on the image and press Ctrl+C.

Frequently asked questions

Do screenshots and recording cost anything?

No. Screenshots, annotation, text extraction, screen recording and format conversion are all free in the 77Fox desktop app. Only screenshot translation through 77Fox Cloud needs a paid plan and uses AI credits; the free channel, your own key, local models and local CLI tools use no 77Fox credits.

Are my screenshots uploaded?

No. Text recognition runs on your PC and screenshot pixels never leave it. When you click Translate, only the recognized text goes to the engine you picked, and with a local model even that stays on your PC.

Does it work offline?

Screenshots, annotation and image conversion all work offline. Text extraction works offline once its recognition model is downloaded, and screen recording, video conversion and HEIC photos once ffmpeg is installed. Translation works offline only with a local model.

Is there a watermark or a time limit on recordings?

There’s no watermark: 77Fox writes only the screen, window or region you chose into the MP4. The file grows the longer you record, so leave enough free disk space before a long session. Files go straight to a folder on your PC and are never synced or uploaded.

Can I record in 4K or at high frame rates?

You can set them, but what you actually get depends on your graphics card and the encoding path: on our test machine, GPU encoding at 1440p and 4K allowed at most 30fps. For high frame rates use 1080p and record a 10-second test first.

Does restoring an NCM file change the sound?

No. Restoring the original format takes the MP3 or FLAC inside the file as it is, without re-encoding, so the audio is identical. Only use it on files you have the right to use, for personal use.

Give it a try.

Bring your most-used prompts in first, then connect your AI agent.