PIY Video Studio

What it is

PIY Video Studio lives in the sidebar under Communication. It is for short, professional videos: a personal introduction, an Upwork or job-application video, a product demo, a tutorial, a client walkthrough or an internal announcement.

The studio has eight sections: My Videos, Record Video, Screen Recorder, Video Editor, AI Video Assistant, Templates, Shared Videos and Settings.

Recording

1. Choose what to record: Webcam, Screen (plus your microphone), Screen + webcam (your face in a bubble) or Audio only.
2. Pick your camera and microphone, the size (720p or 1080p), the frame rate and the quality. Choose No microphone if you only want the screen.
3. Press Start preview. Your browser asks for permission. Nothing is recorded yet.
4. Tick the consent box, then press Record. A countdown runs first. You can pause, resume, mute, turn the camera off and stop. Space pauses and resumes.
5. Review the take. Save, Save and edit, Retake or Discard.

The browser shows its own recording indicator while the camera, microphone or screen is in use, and the studio shows Recording and Sharing your screen on the preview. When you stop, everything is released.

Live effects

On the Look, Audio and Layout tabs you can change how you look and sound. These are live: they are part of what is recorded, and you can change them while recording.

Background. Choose Original, Blur (with a strength slider), a built-in scene (office, studio, grey, slate, dark, blue, sunrise) or My image. A small segmentation model separates you from the background; Edge softness controls how soft the outline is. The model downloads the first time (and only then), runs on your graphics chip when the browser allows it and on the processor otherwise, and the panel shows which and how long each frame takes. If it cannot load you see a notice and the camera shows without the effect. The edge is a soft cut-out, not a hair-level matte, so busy backgrounds or loose hair can show a halo.

Look. Presets (Natural, Studio, Warm, Cool, Cinematic, Bright), brightness, exposure, contrast, saturation, temperature, portrait softening, a filter, zoom and left/right/up/down framing, and mirroring. Hold to compare shows the camera without effects (preview only). Safari has no canvas filters, so colour settings use a simpler approximation there. Sharpness is not offered because a browser cannot sharpen a live camera reliably.

Audio. The meters show the microphone before and after processing, and Listen lets you hear either one (use headphones). Controls: microphone gain, noise suppression, echo reduction and automatic volume (the browser's own processing; the panel shows whether the browser actually applied each), voice isolation (only where the browser provides it), a low-rumble cut, voice clarity with level evening, and a quiet-room gate with an input-sensitivity slider. This is not machine-learning noise removal: steady noise is reduced, but speech from other people or sudden noises are not removed live. Use Smart cleanup afterwards for those.

If something goes wrong

• A blocked camera, microphone or screen shows a message that says how to allow it.
• If the camera or microphone is unplugged, recording continues and you see a warning. If you press the browser's Stop sharing, the recording is stopped and saved up to that point.
• While recording, pieces are also kept in your browser. If the tab crashes or you close it by mistake, open Record Video again and use Recover.
• Uploads go up in pieces and resume if your connection drops: press Save again.

Editing

Open a video and choose Edit. The original recording is never changed. Your edit is saved automatically as instructions, and each export becomes a new video.

• Clips: split at the playhead, delete, reorder, trim by time, set speed (0.5× to 4×), per-clip volume and a dip-to-black transition. Mark in/out (I and O) to remove or keep a stretch. Undo and Redo work for every change.
• Titles & text: an intro and outro title card, and text on the video with its own time, position, size and colour.
• Branding & music: a logo in any corner, and background music from a file you upload. Only use music you have the right to use.
• Audio: reduce noise, enhance the voice, normalize loudness, fade in and out, correct sound that is out of step with the picture, and find silent gaps. Render audio preview lets you hear before and after.
• Transcript & AI: transcribe, search, correct, and jump to the moment. Select sentences, filler words, repeats or long pauses to remove them; the studio always shows what will be removed and waits for you to confirm. Your corrections never overwrite the original transcript text.
• Export: MP4, WebM or audio-only M4A; 480p to 1440p; landscape, vertical or square; optional burned-in captions. A progress bar follows the job on the server, and you can cancel or retry it. Render quick preview makes a small version with titles, logo and captions applied.

The preview in the editor plays your clips in order and shows text and captions. Title cards, the logo, music and audio clean-up are applied when you export or render a preview.

Editing works best on a larger screen. On a phone you can record, watch, trim, cut and export.

Smart cleanup

In the editor's Audio panel, Smart cleanup finds filler words, repeated words, stutters, false starts, long pauses, dead space at the start and end, loud bursts, sudden volume changes, possible breaths and noises, and frozen picture. Choose a sensitivity (cautious, balanced, thorough), press Analyse, then tick the items to remove. Every item shows its time, why it was flagged and a confidence. Low-confidence items (for example a possible breath, which is a guess from the sound alone) are never pre-ticked. Loud bursts and volume changes are marked for review rather than cut, because cutting them could remove speech. Accepted items go through the normal Review before removing step, can be undone, and never change your original recording.

PIY Bot

In AI Video Assistant you can paste a job description or client brief, add the facts about yourself that you approve, and ask for a script of a given tone and length. You can then improve the opening, shorten, expand, change the tone or ask for talking points. The script appears as a teleprompter while you record, with adjustable size, speed and a progress readout.

PIY Bot only uses the facts you gave it. Where it needs one you did not supply, it leaves a [bracketed placeholder] for you to fill. Review everything before you record: it cannot check that your facts are true.

Ask PIY Bot to edit (top of the Audio panel) turns a request such as "remove filler words but keep my message" or "prepare this for a client presentation" into a short plan using the editor's real operations: smart cleanup, noise reduction, voice enhancement, loudness and fades. Each step has its own button; nothing happens until you press it, and cleanup items are still reviewed one by one. Things that cannot be done to an existing recording (changing the background, brightness or colour, changing what was said) are listed as not possible, with where to do them instead (Record, Look, before recording).

After transcribing, Review with PIY Bot can suggest a title and description, a summary, unclear moments, missing points compared with a brief, wording corrections and editing tips. Corrections appear next to the transcript line marked as suggestions, and nothing changes until you accept it.

An optional live speech preview while recording uses your browser's own speech service, which may send audio to its provider. It is off by default.

Templates

Standard templates cover introductions, Upwork and LinkedIn videos, job applications, client proposals, demos, walkthroughs, explainers, tutorials and announcements. A template sets the recording type, the webcam layout, the suggested length, a script outline and an optional intro and outro title. Everyone can make personal templates. Administrators can make templates for the whole organization.

Sharing and privacy

• A video is private to you (and administrators) until you share it.
• Share it with chosen people or the whole organization. They watch it inside PIY Action after signing in; they cannot download or edit it. The link only works for people who may watch.
• Videos are streamed through the server with permission checks. They are never in a public folder.
• Deleting a video removes its file, thumbnail, transcript and edit. Edited exports are separate videos and stay until deleted. Unfinished uploads are removed after a day, and temporary processing files are deleted when each job ends.
• Transcripts and scripts go only to your organization's speech and PIY Bot services, and only when you ask.
• Only record yourself, or people who have agreed. Do not record sensitive screens you are not allowed to share.

What the server needs

For administrators. Recording, upload, playback, sharing and templates work without extra software. These need setup on the API server:

Optional limits: STUDIO_MAX_MB (largest recording, default 2048), STUDIO_JOBS_CONCURRENCY (default 1), STUDIO_JOB_TIMEOUT_MIN (default 60), STUDIO_FFMPEG_THREADS (default 2). Videos are stored under STORAGE_DIR/studio.

Settings in the studio shows which of these are available right now.

Limits

• Speakers are not identified; the speech service returns text and times only.
• Exporting to outside services is not connected. Download the file or share it inside PIY Action.
• Screen recording needs a desktop browser. Phones can record the camera and microphone.
• Background blur and replacement are best on a computer with a graphics chip; they are optional.
• The editor has a single sequence of clips with text, logo and music. A multi-track timeline (several video and audio tracks, keyframes, waveforms) is not built yet.
• Live noise removal uses the browser's own suppression plus a filter chain, not a trained voice-isolation model.

AI noise removal (new)

On the Audio tab, AI noise removal runs your microphone through RNNoise, a small neural network that separates speech from fans, keyboards, traffic and room noise. It runs on your device (an AudioWorklet at 48 kHz, about 10 ms delay) and is part of what is recorded. A strength slider mixes the processed and the original sound. While it is on, the browser's own noise filter is switched off so the model hears the untouched microphone. The panel says Running, Loading or Not available (for example in a browser without AudioWorklet); it never claims AI processing when it is not running. In an automated test with synthesised speech over fan, hum and keyboard noise, the noise between words dropped by about 38 dB while the speech level was kept. That test is not a listening test with real rooms and voices.

Multi-track editor (new)

In the editor press Multi-track editor (a video already edited this way opens there). Your recording and uploaded files are never changed; the project stores only edit decisions, and an export renders a new video from them.

• Tracks: several video, audio, text and image tracks. Mute or hide a track, remove it, add more.
• Clips: drag to move (also to another track), drag the edges to trim, split at the playhead (S), delete (or delete and close the gap), duplicate, extract a clip's sound onto its own track, freeze a frame, speed 0.5×–4×, volume, fade in/out, volume ducking, per-clip noise reduction and voice improvement.
• Picture: brightness, contrast, saturation, temperature, black and white; on upper video and image tracks size, position and opacity (picture-in-picture).
• Text: plain text, titles, lower thirds and subtitles with position, size and a fade or slide animation; title cards inserted at the playhead.
• Library: add audio, image or video files you uploaded as assets.
• Waveform of the recording on its clips; zoom; undo/redo (Ctrl+Z / Ctrl+Y); keyboard shortcuts.
• Export profiles: 480p to 1440p, 24–60 fps, MP4, WebM or sound only, shape (wide, tall, square) and burned-in subtitles; progress, cancel and retry. Preview render makes a quick 480p file.

The preview player shows your recording at the playhead. Other tracks, titles and looks are not composited live; they appear in the rendered preview and the export.

Smart Cleanup and PIY Bot

Cleanup transcribes the video with the built-in speech recognition (English, Whisper tiny, runs on the server; or your configured speech service), then lists long pauses, filler words, repeats, stutters, dead space and sound problems, each with a time, reason and confidence. Untick anything you want to keep, then remove the rest in one step; undo brings it back. It can miss a filler the recogniser did not hear (it sometimes writes "uh" as "a"), so review the list.

PIY Bot turns a request such as "remove long pauses and filler words, add subtitles and improve my voice" into steps with exact timestamps. You tick the steps you approve; they run through the same editing operations as the manual editor and can be undone. Background replacement cannot be added to a finished recording; choose a background on the Look tab before recording.

Known limits

• The multi-track preview is not a live composite; use Preview render.
• Background removal and blur are live-recording effects only.
• Smart Cleanup speech analysis is English only with the built-in recogniser.
• Crossfade transitions between clips are not offered (fades in and out are).

Choosing your camera and microphone

On Record Video → Source pick a Camera, a Microphone and (where the browser allows it) a Speaker for the Listen monitor. The lists show the devices your browser actually finds, with their real names; before you allow access some browsers only give generic names, so names appear once the preview starts. Refresh devices re-reads the list, and it also updates by itself when you plug in or unplug a camera or headset. The Devices box shows what each device is really delivering (for example 1280×720 at 30 fps) and whether it is live; speak and watch the level meter on the preview to test the microphone.

You can switch camera or microphone before and during recording. A switch keeps the recording running: the new camera feeds the same picture pipeline, so background replacement, blur and colour settings carry over, and the new microphone feeds the same audio chain, so AI noise removal, voice enhancement, gain and the gate carry over. There is still only one audio track in the file. There is a very short moment while the new device starts. If the new device cannot be opened, the old one keeps working and you see why. If the device you are using is unplugged, recording continues without it and you are warned; choose another device from the list.

Device ids can change between visits, so the saved choice is not reused across sessions: the list always shows what is connected now. Speaker choice depends on browser support (Chrome and Edge).