Turn your dot
into a video editor.

Set up a video-editing project for me. Start on your cloud computer if it can support the work.

First report where the job will run. Check FFmpeg/ffprobe, Python, faster-whisper, Node.js, HyperFrames and its browser renderer, available memory and disk space. Separate installed tools from things you have actually tested.

Use the current official documentation. Prefer local Whisper transcription on that computer so we do not need a transcription API key. Choose a model and compute type that suit the hardware; try turbo if practical and explain any smaller-model choice. Check that the model really downloads and loads.

For HyperFrames, use an available supported integration or set up its CLI in the project. Do not assume locally installed skills are available in a cloud task. Test the renderer, font loading and audio.

Create a separate project folder. Preserve my original recording. Tell me about any required permission or paid service before using it. Do not change security settings to make a download work.

With my supplied clip, render a five-second test containing video, audio, one caption and a simple visual. Return the actual MP4, the project location and a short passed/blocked report before editing the full video.

Audience: [who is watching]
Main point: [what they should understand]
Platform: [Reels / Shorts / TikTok / other]
Target length: [approximate length or keep the whole message]
Original video: [attached file or accessible location]
Names and terms: [exact spellings]
Keep: [lines, pauses or moments that must stay]
Remove: [obvious failed takes or mistakes]
Captions: one word at a time, lowercase, no punctuation
Font: Inter Medium 500; start at 84 px on a 1080 x 1920 canvas
Caption color: white, with a 3 px black outline
Visuals: [what would help explain the idea]
Opening image: [file, first frame and end cue]
References: [supplied examples or links]
Music: none unless I supply or approve a track
Delivery: MP4, editable project and change notes
Access: keep private until I approve sharing

Edit my talking-head recording using the brief and assets above. First confirm the setup test passed and that you can read the actual video.

Keep the original untouched. Work in a separate project folder. Save numbered versions, editable source files and a short change log.

1. Transcribe the actual speech with word-level timestamps. Check names, numbers and unclear phrases against the audio. Keep a readable transcript and a machine-readable timing file. Flag uncertainty.

2. Make a conservative rough cut. Remove clearly failed takes, repeated starts and distracting dead air. Keep natural pauses, complete word endings and the intended meaning. Keep audio and video together. Save a source-to-edited-time cut map, then remap caption and visual timings. Show the rough-cut MP4 for review.

3. Add one-word-at-a-time captions. Start with lowercase, no punctuation, Inter Medium 500 at 84 px on a 1080 x 1920 canvas, white with a 3 px black stroke. Keep the caption position consistent and clear of the face and platform controls. Preserve proper names and punctuation in the separate transcript.

4. Add a small set of visuals that explain the speech: diagrams, a simple timeline, comparisons or authentic supplied screenshots. Prefer editable HTML/CSS/SVG/GSAP rendered with HyperFrames when supported. Record asset sources. Follow my opening-image timing exactly.

5. Render a short sample covering a cut, fast speech, a long caption word and one visual. Let me approve the style before the full render. Do not add background music unless I approve it.

6. After approval, export 1080 x 1920 MP4 with H.264 video and AAC audio. Use 30 fps for this preset, checking source frame rate, rotation and HDR/color handling first. Reframe without stretching.

7. Check the exported file for clipped words, caption mistakes, timing drift, missing assets, black frames and audio/video sync near the beginning, after cuts and near the end. Separate checks you ran from anything I still need to watch.

8. Deliver the MP4, editable project, corrected transcript, word timings, cut map, asset sources and concise QA notes. If delivery needs a smaller file, make a measured review copy and retain the better export.

Keep the footage and review private. Ask before public publishing or using a paid service. Show concrete outputs and explain blockers plainly.

Transcribe my recording with word timestamps and check the glossary in my brief. Save the corrected transcript and timings.

Make a conservative rough cut. Remove clear mistakes, duplicate attempts and distracting gaps. Preserve the full wording, useful pauses and complete word endings. Listen across each join.

Save each retained segment's original start/end and edited start/end in a cut map. Remap all word timings to the edited timeline.

Return the actual rough-cut MP4, cut map and any ambiguous take choices. Stop before adding visuals so I can review the spoken cut.

Using the approved rough cut, render a short caption sample with fast speech, short words and a long word.

Show one word at a time, lowercase, without punctuation. Use the actual Inter Medium 500 font at 84 px on a 1080 x 1920 canvas, white with a 3 px black stroke. Keep the position consistent and avoid my face and the target app's controls. Scale the preset if the canvas differs.

Check that timings follow the edited audio and the correct font and weight loaded. Keep the readable transcript separately. Return the actual sample MP4 for phone-size review before the full render.

Propose a small set of visual moments for the approved cut. For each, give the spoken cue, edited start/end time, purpose and asset source.

Use authentic supplied or official assets for product logos and screenshots. Mark explanatory diagrams or illustrative UI as illustrations. Do not invent product screens or factual claims.

Build the approved moments as editable HTML/CSS/SVG/GSAP compositions, using HyperFrames when the renderer is working. Make the animation seekable and deterministic. Wait for fonts, images and video assets before frame capture.

Keep motion restrained, preserve the approved dialogue and leave the caption area clear. Render one sample with its real audio before the full composition.

Use my supplied opening image fully visible from frame 0, with no fade or entrance animation. Keep it until the start of the spoken phrase [exact phrase] in the approved edited audio. Confirm that cue's edited timestamp before changing it.

At that cue, switch to the existing [named animation]. Keep all other approved audio, captions and timing. Verify the first frame and the frames around the switch.

If a supported private hosting route is available, make a private review page for this edit. Put the latest version first and retain earlier versions with change notes and correct download buttons.

Verify playback, access and download paths for the intended viewer on desktop and mobile where possible. Tell me which checks you completed and which need me. Keep sharing private; ask before making it public.

If private hosting is unavailable, deliver clearly named MP4 versions through our existing file workflow. Save the editable project and originals somewhere durable.

In [version], the word ending before the cut at [edited timestamp] sounds clipped. Compare it with the original, restore enough linked audio and video to keep the complete word, and remap later caption and visual timings.

Keep the rest of the approved edit. Save a new version and check the repaired join at normal playback speed.

Use [version] as the approved baseline. Change only [specific element] at [edited timestamp or exact phrase].

Desired result: [what should look or sound different].
Keep: [approved cuts, audio, caption style and other visuals].

Save a new version. Compare the affected section and a later unchanged section before delivery. Summarize the change.

Create a review copy below [actual upload limit] with a little headroom. Keep the higher-quality export. Preserve readable captions and clear speech. Measure the final file size and verify its duration, playback and sync before delivering it.

Turn this approved edit into a reusable project template. Separate the footage and content from style settings.

Save the caption preset, visual layout rules, export settings, corrected transcript, word timings, cut map, editable graphics, render steps and tested dependency versions. Add an asset-source list, QA checklist, version history and README.

Use placeholders in the shared template. Exclude credentials, private links, private footage, model caches and temporary render frames. Include fonts or media only when their licenses allow redistribution.

For the next video, use this template and recheck the environment before editing.

Show me the latest completed artifact and what you are working on now. Report: current step, completed output, current blocker if any, and the next concrete test.

If transcription or rendering is running, report observed progress without inventing an ETA. If no export is ready, say so. Do not call the edit finished until an actual MP4 exists and the checks are reported.