Free auto-captions without watermark: EyeStill Studio vs CapCut, Submagic and VEED
Most short-form creators only need two things from a caption tool: accurate word-level timestamps and clean MP4 export. Because traditional editors run speech-to-text on cloud GPUs, nearly every tool now locks auto-captions behind a $12–$39/month paywall or stamps a watermark across the frame. EyeStill Studio runs the entire pipeline on your own GPU inside the browser — free, with no watermark and no sign-up — and anchors each word to the speaker's face.
Why every cloud caption generator added a paywall
When you drop a 60-second vertical clip into Submagic, VEED, or Kapwing, your browser uploads the video file to an external server, runs speech recognition on a server, renders an overlay in a remote container, and streams the finished MP4 back to you. Every export costs the provider server compute and bandwidth, which is one reason free tiers tend to come with limits: Submagic's free plan, for example, is a few watermarked videos a month, and VEED's free plan exports at 720p with a watermark. CapCut's rules for free auto-captions have also changed over time and differ by app version and region, so check the app before you rely on them.
Modern desktop and laptop browsers no longer need a server for this job. With WebGPU and WebCodecs, your local graphics card can transcribe speech with OpenAI Whisper, detect faces in every frame with BlazeFace, composite typography on an HTML5 canvas, and encode a hardware-accelerated H.264 MP4 without sending a single byte over the network.
EyeStill Studio vs CapCut Pro, Submagic, VEED and Kapwing
| Feature | EyeStill Studio | CapCut Pro | Submagic | VEED.io / Kapwing |
|---|---|---|---|---|
| Price for clean export | Free (no watermark, no sign-up) | Pro from about $8–$20 / mo (varies by region and plan) | From $19 / mo ($12 billed yearly) | VEED from about $18–$19 / mo; Kapwing: check its pricing page |
| Free tier watermark | None | Free plan limits vary by app version | Watermark on the free plan (3 short videos a month) | VEED: watermark and 720p on the free plan; Kapwing: check |
| Caption placement | Follows speaker's lips (BlazeFace 60 fps) or custom anchor | Static / manual keyframes | Static bottom or center block | Static bottom or center block |
| Eye-movement design | Single-word ORP red pivot letter (zero eye travel) | Multi-word karaoke bounce | Multi-word pop / highlight | Line-by-line subtitle block |
| Speech-to-text engine | On-device Whisper (WebGPU / WASM) | Cloud ASR | Cloud ASR | Cloud ASR |
| Video privacy | 100% local (files never leave browser) | Uploaded to cloud servers | Uploaded to cloud servers | Uploaded to cloud servers |
Why bottom-third karaoke captions hurt viewer retention
Traditional Reels and TikTok captions sit in the lower third of the 9:16 frame and highlight words left-to-right as they are spoken. Eye-tracking studies of subtitled video find that viewers spend a large share of their time on the subtitle bar, which leaves less for the speaker's expressions, eye contact and gestures.
Single-word rapid serial visual presentation (RSVP) solves the horizontal scanning problem, but only if two conditions are met:
- Spatial colocation with the speaker: if a single-word caption sits at the bottom of the screen, the viewer is still staring at the chin or shirt collar instead of the face. EyeStill Studio runs MediaPipe BlazeFace across the video timeline, smoothed with a Kalman filter so the caption glides with the speaker's mouth without camera-shake jitter.
- Optimal Recognition Point (ORP) alignment: centering words by total character width makes the eye jump slightly as word lengths change. EyeStill shifts each word horizontally so its recognition pivot letter — roughly 30% from the left edge — stays locked to the exact same pixel, marked in vermilion red.
How to caption a Reel, Short or TikTok in 60 seconds
- Open eyestill.com/studio in Chrome, Edge, Brave, or Arc on desktop. No login is required.
- Drop any vertical (`9:16`), square (`1:1`), or landscape (`16:9`) MP4, MOV, or WebM video onto the canvas.
- Select the spoken language. On first run, the browser downloads and caches the quantized Whisper model locally; subsequent videos transcribe immediately on your GPU.
- Inspect the word timeline: edit any misheard word, adjust WPM pacing, or toggle between automatic lip-tracking and fixed focal placement.
- Click Export MP4. WebCodecs renders the burned-in ORP captions at full resolution directly to your Downloads folder with zero watermarks.
What if you want the same captions while watching YouTube?
EyeStill Studio burns captions into exported video files for creators publishing to TikTok, Instagram Reels, and YouTube Shorts. If you are a viewer watching long-form videos, lectures, or podcasts on YouTube, the EyeStill Chrome extension runs the exact same BlazeFace lip-tracking and ORP word drum live over any YouTube video — including real-time subtitle translation — without downloading or re-encoding the video.