Skip to main content
The Video editor is a workspace-gated preview for chat-driven editing. Start with a source video, describe the result you want, and review the cut before daoco produces the final render. The editor uses both transcript and visual evidence. It can bring still images and approved library assets into a video project, but it is not an asset-only slideshow generator: begin with source footage, a screen recording, or an existing library video.

​
Enable the preview

1

Open Settings

A workspace owner or admin opens Settings → Preview features.
2

Enable Video editor

Turn on Video editor and save the workspace settings.
3

Open brand chat

Attach a video, select a library video, provide a source URL, or continue an existing video project in the same thread.

​
What you can make

Caption and align short-form video

Burn in captions, hooks, and timed text. daoco keeps text inside platform-safe areas and avoids covering other important on-screen text.

Clip longer recordings

Find strong moments, remove pauses, tighten pacing, and reorder scenes into focused short-form cuts. Pauses can be removed from footage with no transcript, measured from the audio itself.

Construct clips from assets

Combine source footage with still images and approved library media. Use brand assets as image scenes, overlays, logos, or supporting panes.

Reframe people and screens

Adapt horizontal interviews and screen recordings for vertical placements with subject-aware framing, readable detail, and safe crops.

Build composed edits

Sequence video and image scenes, add image holds, eased zooms, scene transitions, speed changes, music, sound effects, captions, animated branded text, and one- to three-pane layouts.

Create and refine variants

Compare candidate cuts, request a different direction, and iterate from the reviewed edit without rebuilding the project from scratch.
A long edit runs in several passes. Each pass picks up the saved timeline and the work the earlier passes finished, so it goes straight to what is left rather than re-reading every source. The video panel follows the cut being assembled, so a multi-clip stitch stays on screen while the edit continues. The panel can appear before the edit is finished: a saved timeline is not a rendered preview, so review the preview before exporting.

​
Direct the video before editing

Every new video editing request starts with the Assistant Director, including /clip, /edit, and requests to cut uploaded footage. It prepares the brief and asset handoff before the Video Editor starts. Follow-ups on an existing edit, such as trimming a cut, changing captions, or rendering another preview, go directly to the Video Editor. The Assistant Director creates a production container beside the conversation with three views:
  • Brief is an editable Markdown plan for the outcome, audience, story, scenes, text, audio, brand rules, and deliverables.
  • Footage & assets shows the footage, images, logos, graphics, music, and references selected for this video. Each asset has an editorial role and is marked when the editor must use it or when it is reference only.
  • Still needed keeps missing material, unanswered direction choices, rights to confirm, and requests the editor cannot fulfil visible until they are resolved or dismissed.
The director searches your brand library by meaning as well as names and tags. For analyzed videos, it can match a request to a relevant shot and its time range. It checks the current media evidence before it says what footage contains. Ask it to compare directions, suggest existing assets, replace a rejected choice, or identify what needs to be generated or imported. Asset creation and connected integrations can fill those gaps without hiding them from the plan. If no accessible source video is found, the director records the missing footage under Still needed and asks you to upload or select a video. The brief stays blocked until footage is available; it does not represent selected clips or an edit ready for approval.
You can edit the text of any brief section directly in the panel. Saving sends the change through the Assistant Director and records the request in the conversation; the panel shows the edit as sending until the director applies it, and the director reports a conflict instead of overwriting a brief that changed while you were editing. Asset and gap changes stay in the same durable container, so the plan survives a refresh and remains reviewable. Leave an asset out or mark a missing item resolved from the panel, and the director records it. The director prepares the source transcripts and visual evidence, recommends specific moments with timestamps, and lays out the asset order and edit instructions. When the completed plan is ready for review, press Approve brief & start editing. Your approval applies to the version you reviewed. Asking for clips at the start of the thread does not approve a plan you have not seen. In dashboard chat, you can also send a short approval such as approved, approve the brief, or start editing while the current brief is ready for review. This approves that version and starts the same handoff as the button. If you need changes, describe them before approving. Typed brief approval is not available through hosted assistant channels. The Video Editor reuses the prepared sources and executes the approved plan. It owns cuts, previews, and renders from that point forward. Follow-up changes to the existing edit go directly to the editor. It updates the edit brief when needed, applies your changes, and renders a fresh preview without asking you to approve the production brief again. The editor panel opens the cut targeted by an edit. Caption and framing controls show its saved settings; an older preview keeps its original appearance until you render a new preview. Preview ready means you can watch a review copy. Preview checked means that copy passed its quality check. Only Ready to download means a final MP4 exists for the current edit version. If your cut only has a preview, the panel shows the final export’s estimated premium-token cost. Click Export MP4 to approve that cost and start the export directly. You do not need to send a chat message or wait for an agent. When the final MP4 is ready, the button becomes Download MP4. The download keeps your chat open and saves the file with the cut title. If it fails, the panel shows an error and you can retry. Starting export means the backend is accepting your request; Rendering MP4 means it is queued or running. Exporting keeps the completed preview and quality-check stages visible. The agent can see your export in the project activity and render status. If it requests the same export, it reuses that job. You can still ask for an export in chat. That route keeps its paid-render approval step. Starting a new production requires its own brief approval. When you ask for a post with “this video” in a thread with one existing cut, Daoco attaches that cut’s final MP4 and duration. If there are several cuts, choose the intended cut and use Create post.

​
Start an edit

Use /clip or /edit to start video work. A new project begins with the Assistant Director and a production brief for your approval. Once editing has started, requests to trim, change captions, reframe, or render another preview go to the Video Editor in the same thread.

​
Caption and reframe a short video

​
Turn a longer recording into clips

​
Construct a clip with library assets

​
Reframe a screen recording

Include the intended platform, target duration, audience, required moments, caption style, and any assets that must appear. If you do not specify them, daoco proposes an edit brief for review.

​
Working with several videos

Attach as many videos as the edit needs. daoco lists what you attached, opens a separate edit project for each one, and keeps their transcripts and analysis apart — it never quietly edits the first file and calls it done. If a video was already opened in the same conversation, its existing project is reused instead of being transcribed again. To combine them, ask for one cut and name the order:
A stitch plays each clip’s edited cut, not its raw file: pauses you already trimmed stay trimmed, the silence at the start and end of every clip is cut back to a short breath, and the clips are levelled so a phone recording and a studio recording sit at one volume before the final normalization. Joins are a short crossfade unless you ask for hard cuts, and you can trim what each source contributes. A clip with no transcript keeps its edges until you ask daoco to remove the pauses on the stitched cut, which it measures from the audio. Captions cover every stitched video that has a transcript, so a source still being processed contributes no captions; wait for it before finalizing. With many short clips, ask for the best moments across all of them first. daoco ranks the strongest takes from every clip in one pass, and you choose the order:
Clips shot on different cameras rarely match. Ask daoco to match the look and it measures the brightness and saturation of each source and nudges them toward one another. You can also ask for a single look on the whole cut (warmer, more contrast, less saturated) or a correction on one scene. The corrections are small on purpose; the phone-sized preview is where you judge them. The result is an ordinary cut: keep asking for trims, reordering, caption changes, or reframing on it.

​
Put a demo or an image behind you for a segment

Popular product and tutorial shorts cut to a screen recording, a demo clip, or a still while the person keeps talking from a small pane in the corner. Ask for it on the segment you mean, and name the footage or image from your library:
The rest of the cut is untouched. The pane keeps clear of the platform’s controls (it moves to another corner if it has to), and it fades in and out. An image works the same way as a video, so a product shot or a slide can fill the frame for a beat. B-roll that should replace you entirely for a moment is a plain cutaway; say which you want.

​
Change the background or steady the shot

Footage shot on a green screen keys cleanly: name the screen colour and what should show behind the person, a colour, an image, or a video from your library. Footage shot anywhere else can have its background removed by a matting model, then the same choice of backgrounds goes behind the subject. That runs once per source, is reused by every later cut of the same footage, and uses premium tokens per 30 seconds of footage (more on the quality tier, which takes clips up to 30 seconds). daoco tells you when the model is still working and finishes the cutout when you ask again.
Shaky handheld footage can be steadied on the whole cut or one scene. It is free and crops in slightly to hide the corrected edges, so preview it.

​
Add a voiceover

Explainers, listicles, and product walkthroughs often need narration the footage does not have. Give daoco the script and, if you like, a voice, and it reads it into a narration track, files that track in your library, and mixes it over the cut. The music bed dips under the narration on its own; ask for the footage’s own sound to dip too when it competes.
Recorded your own? Attach the MP3, WAV, or M4A in chat and say it is the voiceover, or point at an audio file already in the library; a recording you supply costs nothing. daoco tells you when a narration runs past the end of the cut so you can extend the cut or tighten the script. Each synthesized read uses one premium token.

​
Freeze frames and name cards

Hold a frame on a reaction, a reveal, or a “wait for it” beat by asking for a freeze at a time; it holds still for as long as you say, silent, and the cut resumes after. Introduce a speaker with a lower third: a small name-and-title card in the lower band of the frame that stays clear of the captions.

​
Cut to the beat

For music-driven montages, set the bed first, then ask daoco to find its beat. It measures the tempo and beat times, and from then on the cuts between scenes can snap to the beat, and zooms or sound effects can land on the hits.
If a track has no steady tempo, daoco says so instead of guessing where the beats are.

​
Caption styles

Burned-in captions use the transcript. If it mishears a brand, product, or person’s name, ask the editor to correct that word and render a new preview. The correction applies to every occurrence in the cut, including stitched scenes, without changing the source transcript.
Ask for a style by name, or describe the look you want:
  • Karaoke highlights each word as it is spoken, on bold display type. The usual choice for spoken short-form.
  • Bold pop is larger and centered, for high-energy cuts.
  • Boxed puts a solid box behind the caption lane so text stays readable over busy or bright footage.
  • Minimal keeps a single narrow line when the frame has to stay visible.
  • Regular is plain white outlined text with no styling.
Your brand colors and fonts fill in whatever the style leaves open, so captions match the brand without you asking. You can still name a caption color, highlight color, box, or font, and daoco keeps captions clear of faces, on-screen text, and platform controls.

​
Movement and pacing

Motion is applied sparingly, one treatment per clip, so a cut still reads as edited rather than decorated.
  • Zooms ease in on a beat, hold, and ease back out. Say where the punch should land (“push in on her face when she says it”) and how long the whole move should last; the hold sits inside that time.
  • Transitions between scenes can be a crossfade, a dip to black, a slide, or a zoom. A transition overlaps the two scenes it joins, so the clip gets slightly shorter than the sum of its scenes.
  • Speed changes the pace of a span: speed up a walk-in or a long UI wait, slow down a moment worth landing. Captions stay in sync automatically.
  • Text and asset animation is a short fade, pop, or slide as a caption card, hook, logo, or overlay enters and leaves. Where the text settles does not change, so safe areas still hold.

​
Music, sound effects, and stickers

daoco keeps a licensed media catalog so you never have to go find a track. Ask for the feel you want and the editor searches it for you.
  • Music beds by mood and energy — upbeat, cinematic, corporate, chill, electronic, luxury, energetic, emotional. The bed ducks under speech and fades in and out on its own.
  • Sound effects for a moment in the cut: whooshes, clicks, pops, impacts, risers, transitions, notifications, and room ambience.
  • Stickers — arrows, badges, shapes, and frames — placed anywhere on the frame, with the same safe-area rules as any other overlay.
Your own logos, product shots, and footage still come from Assets; ask for “our logo” and the editor places the brand’s core logo without you naming a file. When nothing in the catalog matches what you asked for, daoco says so rather than substituting something else.

​
Why this cut

Every candidate carries a score out of 100 and up to three short reasons for it, shown on the card and under Why this cut. The score rates five things independently — how well the opening earns attention, how clear the moment is without context, whether it pays off, how tight the pacing is, and whether there is something worth watching — and the hook number is shown on its own because it is the part you can act on before rendering. It is an estimate daoco makes from the transcript and the frames it sampled, not a prediction of views. Use it to compare cuts and to see what a weak candidate is weak at. After you ask for trims, silence removal, or a new hook, ask daoco to re-score the cut so the number describes the cut that now exists.

​
Multi-speaker recordings

For interviews and podcasts, daoco reads who is speaking from the transcript and reports how long each person talks. Tell it who they are and it uses the names from then on:
Once speakers are named you can also ask for a colour per speaker in the captions, and for the framing to follow whoever is talking. Speaker-following needs the speakers to be distinguishable on screen; where they are not, daoco keeps the largest face in frame and says so instead of guessing.

​
Review the edit

The Video editor panel keeps the cut and its review state beside the conversation. For full-frame footage inside a vertical canvas, you can request either blurred padding or a solid black background. If the editor finds a preview does not match the approved brief, it records a hold with the reason. The panel shows that hold and blocks final export until a corrected preview passes review.
  • The strip at the top of Review shows where the cut is on its way to a finished video (cut, preview, checked, final), what is happening to it right now, and what you can do next. Final renders use premium credits, and the approval bar shows the exact number before you approve.
  • Review shows the active cut, social preview, captions, crop, duration, score, rationale, QA status, and cut timeline.
  • Compare shows candidate cuts so you can choose a direction or ask for another pass.
  • Render this cut creates a representative preview when the current candidate has not been rendered yet.
  • Request changes keeps the same project active for trims, new scene order, caption changes, zooms, hooks, music, or reframing.
  • Export MP4 shows the estimated premium-token cost and starts the final render when you click it. The button becomes Download MP4 when the export is ready.
Previews render at half the delivery size so they come back quickly. They use the same composition plan as the final, so the framing, panes, overlays, captions, and safe-area placement you review are exactly where they land in the delivered file. Loudness is set on the final render, not on the preview.

​
Adjust

Under the cut, the panel shows the saved caption style, format, and voiceover. Ask for edits in chat. You can request caption changes, reframing, trims, music, a new score, or a fresh preview. Name the cut when several are in the project. Custom text can appear for a specific time range. For example, ask: “On this cut, show ‘Three things to know’ from 0:02 to 0:06.” The editor places this separately from transcript captions. Ask for a new preview to review it in the video. On Compare, an unbuilt moment has a Build this moment button that asks the editor to turn that suggestion into a reviewable cut.

​
Speakers

For a diarized recording, Speakers lists each voice with how long they talk. Choose Name to tell the editor who someone is; it uses that name from then on, and it is what caption colours and speaker-following key off.

​
Credits

If a cut uses music, sound effects, or stickers from the licensed catalog, a Credits line appears under the delivery buttons with the attribution each track needs. The download hint says so too, so a file never leaves uncredited.

​
Deliver the finished video

After the final render:
  • Download the MP4.
  • Ask the editor in chat to save the video to Assets for reuse.
  • Choose Create post to hand the finalized video to the social media manager with its placement context intact.
See Create posts for the post review and publishing flow, or Assets for library reuse and downloads.

​
Current limits

  • Video editor must be enabled for the workspace.
  • A video source is required to start a new project. When more than one video is attached and it is not clear which one an edit is for, daoco asks instead of guessing. Images and library assets can support or compose the edit after the source is available.
  • Every new edit normalizes its finished mix to a consistent loudness, and a music bed fades in and out unless you ask for a hard cut. Ask in chat to change the loudness target, quiet or mute a specific stretch, or add a closing fade.
  • One pane owns the primary audio in a multi-pane composition. Supporting panes remain muted unless the edit explicitly replaces the primary audio.
  • Subject-aware reframing can fall back to a non-destructive fit when reliable subject detection is unavailable. Following the active speaker additionally needs speaker labels on the transcript and speakers who are distinguishable on screen; otherwise it follows the most prominent face.
  • Candidate scores are estimates from the transcript and sampled frames. They are not audience predictions, and they describe the cut as it was when it was scored.
  • A preview is the visual review artifact; a saved edit plan alone is not proof that captions, crops, or overlays render correctly.
  • The editor can move a logo or text slightly to clear platform controls, then render and check a fresh preview without another brief approval. It preserves the approved asset and timing. Explicitly locked positions and substantial design changes still require your direction.
  • When you ask for the brand logo without selecting a file, the editor uses the current primary logo from brand settings. Its default position clears the selected platform’s controls, and the image keeps its proportions.
  • Requests beyond the supported clip or render limits stop with an error instead of silently dropping part of the edit.
  • Background removal is a model matte, not a hand-drawn mask: fine hair and fast motion can show soft edges, and there is no per-frame correction.
  • Colour corrections are bounded to matching and gentle looks; heavy grades and LUTs are not supported.

​
What uses premium tokens

  • Previews are free. Each cut gets three finished previews per edit revision; change something about the cut to start a fresh allowance.
  • Final renders charge on delivery: 2 tokens for the first output of a source up to an hour (plus 1 per additional hour), then 1 per additional distinct output. The approval bar shows the exact number before you approve.
  • Subject tracking (keeping a face or the active speaker in frame) charges 1 token per started minute of source footage, once per source, whether or not a subject is found. It is included in the approval estimate.
  • Voiceover synthesis charges 1 token per read. Loudness and colour measurements, beat detection, picture-in-picture layouts, chroma keys, and stabilization are free.
  • Background removal charges 1 token per started 30 seconds of source on the fast tier and 3 on the quality tier, once per source; later cuts of the same footage reuse the result.
  • Pause detection on a source without a transcript charges 1 token once per project. Sources with a transcript find pauses from the word timings for free.
See Understand token usage.