VideoMode
Plan, generate, and render videos from scripts, assets, and templates — with storyboarding, QA, and a render queue.
VideoMode is a focused workspace for building short videos inside the desktop app. You bring a script, idea, reference link, or source assets; pick a template; let the agent storyboard and plan each scene; then generate clips, music, narration, captions, timeline edits, and handoff packages before previewing and rendering the final cut.
Open VideoMode from the left sidebar.
Start a Project
- Open Video from the sidebar.
- In Choose what to make, select Video, From template, or From URL/file/folder.
- Add a starting idea if you want the project dialog to open with a prepared brief.
- Click Start, or use Create an HTML video when you want an HTML/Motion project.
- Pick a workflow or template. You can start from a classic project template such as Product reel, Explainer, Slideshow, Podcast, UGC ad, or Custom, or use an HTML/Motion template when you want a frame-based animated video.
- Add a project name, optional script or brief, and any reference assets.
- Click Create project.
Each project is stored in your local workspace with its script, source assets, scene plan, generated media, render outputs, and QA reports.
You can also use the same entry panel to jump to DesignMode for Design, Image, or Audio projects without leaving the creative workspace.
The Five Workflow Steps
VideoMode guides each project through a stepper:
| Step | What you do |
|---|---|
| Brief | Refine the script, attach source assets, choose a template and brand kit |
| Storyboard | Generate or arrange scenes in the scene sequencer |
| Plan | Decide how each scene is built — existing asset, AI clip, B-roll, image, lip-sync |
| Generate | The agent runs the plan: TTS, music, image, video, and proxy generation |
| Preview | Watch the assembled cut, run QA, and queue a final render |
You can return to any step at any time. Selecting a step only changes which panel you see — the project state is shared.
Above the editor, the creative workflow header shows the broader project path: Intent, Assets, Plan, Generate, Review, and Export. It shows how many assets are in the project, how many were generated, which steps are ready or complete, and the next recommended action. Click a workflow step to jump to the matching editor panel; choosing Assets opens the asset rail.
Brief
The Brief workspace uses the right rail for project context. Depending on the current selection, the rail can show Brief, Assets, Sources, Brand, Transcript, and Inspector tabs.
- Brief — paste or refine the spoken or on-screen text, project goal, and core inputs.
- Assets — drag in images, video clips, audio, or document references. Filter the list with the All · Video · Image · Audio chip toggle at the top of the rail. Select project assets when you want the Video agent to use them as context for the next request.
- Sources — browse linked folders and source material that stays outside the project folder.
- Brand — review the brand kit used for colors, typography, and scene styling.
- Transcript — work with transcript-linked selections when the current step supports them.
- Inspector — edit the selected scene or clip when a selection is active.
The Assets rail uses the shared creative asset browser. Search by name, filter by kind, switch between grid and list views, select multiple assets, preview files, place media on the timeline, download or delete project files, and mark assets as context for the next Video agent request.
Click Add folder in the Assets rail to attach a folder from your computer. The desktop app opens the system folder picker, asks the workspace to grant access, and then mounts every supported file in that folder as a linked asset — without copying anything into the project.
A brand kit defines the colors, typography, and look-and-feel applied across scenes. You can swap brand kits without re-generating assets.
You can also paste links into the Video agent composer:
- Paste an article or GitHub link when you want the agent to use it as reference context.
- Paste a YouTube link when you want to use that video as source footage. VideoMode downloads it into the project as a normal video asset so the agent can place it on the timeline. YouTube imports are marked as unverified source footage, so review rights and attribution before exporting or sharing.
- Pasted web, image, YouTube, and Vimeo links show preview cards after the message is sent. Open a website or image card to check the source, or preview a video card inline without leaving the project.
Source Review and Auto-Cut
Use the Sources tab when you want to turn longer footage into a cleaner edit.
- Open the Sources tab in the right rail.
- Upload a source video, paste a source URL, or attach source footage from your workspace.
- Confirm that you have the right to import and edit the source.
- Click Analyze to create a transcript and find suggested cuts.
- Review the suggestions. VideoMode looks for long pauses and common filler words, but every suggestion remains reviewable.
- Click Accept cuts when you are ready to apply the approved cut plan to the timeline.
Source analysis can use local speech recognition or a configured cloud speech provider. If cloud transcription would cost money or send audio to a provider, VideoMode asks for approval before using it. When reliable word timing is unavailable, cut suggestions stay in review mode instead of being applied automatically.
The Transcript tab supports text selection. Select a phrase in the transcript, then ask the Video agent to work on that moment. If the source has word timing, VideoMode sends an exact word-timed selection. If it only has scene timing, VideoMode sends an estimated time range.
Generated Captions
Ask the Video agent to generate captions when your timeline includes spoken-word video clips. VideoMode transcribes the clip audio, creates time-synced caption cues, and uses those same cues in preview, render, sharing, and editor handoff.
- Open the Video agent dock.
- Ask it to generate captions for the project, one scene, or one clip.
- Add a style if you want a specific look, such as classic, tiktok-word, hormozi-bold, or karaoke.
- Open the Transcript tab to review the generated caption lines.
- Click a timestamp or word to jump the playhead to that moment.
- Double-click a caption line to fix the text inline.
When you edit generated caption text, VideoMode keeps the cue duration and redistributes the word timing so word-by-word animations still line up. If you move, trim, split, or rebuild video clips, generated captions retime to the source speech when possible. Captions whose source speech was removed are dropped instead of drifting out of sync.
In Preview, use the captions render mode to turn captions off, burn them into the video, or export an .srt sidecar.
Video Agent Dock
The Video agent dock stays with the project. Its conversation history is saved per project, so the agent can remember earlier turns, pasted links, selected assets, and decisions when you come back later.
Use the model picker in the dock header when you want a different Claude model for the current video project. This does not change your default agent runtime for other tasks.
Storyboard
The scene sequencer lays out each scene as a card.
- Click Generate storyboard to let the agent draft scenes from the script.
- Drag scene cards to reorder them.
- Click a card to open the Scene Inspector for that scene.
- Use the inspector to set scene title, on-screen text, captions, narration voice, and timing.
- Add, duplicate, or remove scenes as needed.
Plan
The Plan view assigns a generation strategy to each scene:
| Strategy | Use it for |
|---|---|
| Existing asset | Use a source video or image already in the project |
| AI clip | Generate a video clip from a prompt |
| B-roll search | Pull a matching stock clip from a connected provider |
| AI image | Generate an image (optionally with reference images and seed) |
| Image pan | Animate a still image with a Ken Burns crop |
| TTS narration | Synthesize a voiceover from text |
| Lip-sync avatar | Animate a talking avatar from narration |
For each scene you can also configure music, captions, and a proxy for fast preview.
For AI image and AI clip scenes, the Generate media panel keeps the scene prompt, references, provider, model, aspect ratio, duration, and regenerate controls together. This makes it easier to adjust one scene without rebuilding the whole project.
Generate
The Generation queue runs approved scenes. The agent picks providers based on your settings:
- Music — ElevenLabs Music, Stable Audio, or MiniMax Music
- Narration — Kokoro, ElevenLabs, or OpenAI TTS
- Sound effects and ambience — short SFX or background ambience to drop onto the sound-effect track
- Image — your configured image generators (see Media Generation)
- Video clips — connected video providers, or Hedra for lip-sync avatars
- Proxy — a low-resolution preview generated locally with ffmpeg
You can re-run a single scene without rebuilding the rest. You can also ask the agent to transform an existing audio clip in place — clean it up, extend it, remix it, or swap the narration — without regenerating the whole scene.
Project Flow
For HTML/Motion projects, the Project flow view shows how the brief, source assets, frame plan, render jobs, agent actions, outputs, and exports connect. Use the graph to follow relationships, open generated outputs, review running or failed jobs, and download the debug JSON when you need to inspect the flow more closely.
Preview
The Preview view shows the assembled cut with a scene strip, transport controls, and optional overlays. The viewer header keeps everything you need next to the aspect picker: Share, Re-render, Open output, and Open folder all sit there so you can act on a finished cut without leaving the viewer.
The timeline opens fitted to the full project duration by default. As you move the playhead, the scene strip follows the current scene so the selected scene, inspector, captions, and timeline stay aligned.
Use the playback speed menu next to the aspect picker to review the preview at 0.25x, 0.5x, 0.75x, 1x, 1.25x, 1.5x, or 2x. VideoMode uses the fastest local preview path your system supports and falls back automatically when needed. When a transition has detailed controls, the live preview and hover scrub use those same settings.
When you ask the Video agent to recreate a reference or work from selected assets, it can match the reference aspect ratio, analyze whether each asset should be cropped or kept whole, and suggest a fit before rendering. VideoMode also chooses safer defaults when you drop or attach media: likely logos are shown whole on a white background, while portrait, square, or landscape media that would lose too much content uses a blurred pad instead of being heavily cropped.
| Control | What it does |
|---|---|
| Aspect picker | Preview the cut in 16:9, 9:16, 1:1, or 4:5 |
| Playback speed | Slow down or speed up preview playback without changing the edit |
| Render mode | Local (ffmpeg or Remotion) or Cloud (if a render provider is set) |
| Loudness | Normalize to YouTube (-14), Mobile (-16), or Broadcast (-23) LUFS |
| Auto color | Apply automatic color grading |
| Auto reframe | Re-crop to a different aspect ratio |
| Add to render queue | Submit a final render job |
| Editor handoff | Export a package for a desktop video editor |
Each render is estimated for cost before it runs. If an estimate exceeds your video budget cap, the render is blocked.
Timeline Editing
The timeline lets you make precise edits after the storyboard is built.
| Edit area | What you can adjust |
|---|---|
| Tracks | Video, B-roll, overlay, voiceover, music, sound effect, and caption |
| Clip position | Move clips, trim starts and ends, and align clips to the playhead |
| Cut tools | Split with the razor tool, cut at the playhead, duplicate, delete, or ripple-delete clips |
| Linked clips | Keep related clips together when they came from the same scene |
| Playback | Change clip speed, reverse playback, or mute a selected visual clip |
| Fit | Crop to fill, show full media, stretch, use a background color, or use a blurred pad |
| Filters | Brightness, contrast, saturation, blur, grayscale, sepia, and hue |
| Keyframes | Animate opacity, scale, position, rotation, crop, volume, or captions |
| Visual transitions | Drag transition presets onto adjacent visual clips, then tune duration and direction |
| Audio seams | Smooth or cut audio at clip boundaries |
Use the Select tool for normal clip moves and trims. Switch to the Razor tool from the timeline toolbar, or press B or C, then click a clip to split it at that frame. Press V to return to selection.
Right-click a clip for copy, cut, duplicate, split, delete, and ripple-delete actions. Ripple delete closes the gap left behind by the removed clip. If clips are linked, VideoMode keeps related scene clips together unless you unlink them first.
Use the Playback section in the clip inspector to adjust speed, reverse playback, or mute the selected clip. Audio clips also expose gain plus fade-in and fade-out controls, and those fades are reflected in preview and render output.
Use Keyframes in the clip inspector when you want movement inside a clip, such as a fade, push-in, caption pop, volume ride, or crop reveal. Keyframes are local to the selected clip, so a two-second keyframe always means two seconds after that clip starts.
Open the Transitions tab in the side rail to search the transition library. Drag a transition onto a visual clip edge or the seam between two adjacent visual clips. A badge appears on the timeline; select it to change the transition type, duration, direction, or available Parameters in the inspector, resize it on the timeline, or remove it.
VideoMode only applies visual transitions to adjacent, unlocked clips with enough media on both sides. It automatically limits duration to fit neighboring clips, and it keeps a hard cut when a seam is blocked by a gap, a locked track, or a clip that is too short. Some transitions, such as Clock wipe, Soft wipe, Pixelize, and Polygon iris, expose parameters like angle, feather, center, edge color, sectors, or pixel steps. Local final renders use the browser compositor when needed so those detailed transition settings match the preview. If a selected render path needs a close fallback, VideoMode marks that case.
Hover over the timeline to scrub the preview without moving the playhead. A vertical time label follows your pointer and the preview shows the hovered frame. Hover scrub turns off while playback, clip dragging, or lasso selection is active.
Click a marker on the ruler to open the marker editor. You can rename the marker, move it to an exact time, choose a color, mark it as a chapter, add a comment, or delete it.
The timeline also supports lasso selection, snapping, keyboard shortcuts, track reorder, mute, visibility, lock, and sync-lock controls. Auto-fit runs when a project first opens or when a clip is inserted; after you manually zoom or pan, VideoMode leaves your view alone.
The agent can inspect preview frames before and after visual edits. This helps it verify crop, fit, overlay, caption, and transition changes against what you actually see in the preview, not only against the project plan.
The Video agent can also inspect transition seams, suggest subtle transition changes, remove flashy transitions, or leave straight cuts in place when that is the cleaner edit. Agent transition changes stay reviewable and undoable like other timeline edits.
Audio Editing
VideoMode gives you full control over how every audio clip sounds — voiceover, music, sound effects, and ambience — right on the timeline. Every audio clip shows a waveform so you can see speech, beats, and silence while you trim and align.
Select an audio clip and open the Inspector to adjust it:
| Control | What it does |
|---|---|
| Mute | Silence a single clip without removing it |
| Gain | Raise or lower one clip's volume, in decibels |
| Fade in / out | Set fade durations for a smooth entry and exit |
| Fade curve | Shape each fade — Linear, Equal-power, or Ease-in-out |
| Transcript | View or edit the transcript text attached to a spoken-word clip |
You can also drag the fade handles on the left and right edges of an audio clip to set fades directly on the timeline.
To control a whole track at once, use the track header to set a track volume or mute the track — handy for lowering all your music in one move.
For mixing between clips:
- Audio crossfade — blend two back-to-back clips on the same track so one fades out as the next fades in.
- Audio transition — choose a hard cut or a crossfade at a clip boundary.
- Ducking — automatically lower one track (such as music) whenever another track (such as voiceover) is playing, so narration always stays clear.
- Volume keyframes — ride a single clip's volume up and down over time for custom swells and dips.
The Video agent can do all of this for you. Try requests like "fade out the music over two seconds," "make the voiceover louder," "duck the music under the narration," or "crossfade these two clips." Every audio edit the agent makes is a reviewable action you can undo.
Your audio mix sounds the same in preview, in the final render, and in any editor handoff package — VideoMode applies clip gain, fades, track volume, crossfades, and ducking the same way everywhere.
Render Queue
The render queue appears in the Preview workspace when render jobs exist. It refreshes automatically while jobs run and lets you cancel an active job without leaving the project.
| State | Meaning |
|---|---|
| Queued | Waiting to run |
| Running | Currently encoding |
| Done | Finished — open the output file or share |
| Failed | Render failed — see logs in the project |
| Cancelled | You cancelled the job |
Queued renders run as capacity is available. A project keeps one active render at a time today.
QA Report
After preview or render, VideoMode runs an automatic quality check.
- Black frames — detects unintended dead frames
- Audio clipping — flags peaks above your threshold
- Silent gaps — flags long silences in the cut
- Missing media — warns when a planned asset never generated
- Cut boundaries — checks the area around edits for nearby black frames, clipping, or silence
- Duration mismatch — warns when the rendered file is noticeably longer or shorter than the project timeline
- Loudness peak — reports peak dBFS for the final mix
QA issues appear as a badge on the project header. Open the QA panel to see each issue with a jump-to-scene link.
Templates
Templates let you reuse a project shape — scene structure, default voices, music style, pacing, and aspect ratio.
- Open Settings > Video > Templates.
- Browse built-in templates by Category (Shorts, Explainer, Ad, Tutorial, Product, Podcast, Testimonial, Recap, Announcement, Other) and Pace (Slow, Medium, Fast, Extreme).
- Click a template for details including typical duration, aspect ratios, and pricing range.
- Click Use template to start a new project from it.
From any project you can Save as template to make the current layout reusable.
HTML/Motion templates add a frame-based workflow for animated videos. Open the template gallery to preview the source, fill in template variables, and edit individual frames from the frame strip. Workspace templates can live under your local .neuma/video-templates folder, so teams can keep reusable motion systems next to their projects.
Budgets
VideoMode has its own project budgets. Configure them in Settings > Video.
| Setting | What it controls |
|---|---|
| Max video jobs | Number of render jobs allowed per project |
| Max video seconds | Total generated video duration |
| Max audio seconds | Total generated audio duration |
| Max retries per scene | Repeated failed attempts for the same scene |
| Render cost cap | Hard cap on cloud render spend |
When a project approaches a limit, the project header shows a warning. When a request would exceed a limit, VideoMode blocks it before starting.
Assets Catalog
The project Assets panel is wired to the Assets Catalog, so you can drag any item — local files, AI-generated outputs, cloud media, or stock footage — straight into the timeline or scene plan. Remote items download on demand. Each tile shows a progress badge while it fetches, and you can cancel an in-flight download or retry a failed one without leaving the project. Deleting a project asset that is already on the timeline asks for confirmation and removes the matching clips too.
Use Add assets > Add folder when you want to link a local folder from your computer. In the desktop app, or in the browser build connected to the local API server, VideoMode opens the system folder picker and links the chosen folder as a source instead of asking you to paste a path.
- Hover for a quick preview — hovering a folder card plays its video inline, and hovering an output tile pops up a muted preview anchored next to the row so you can compare renders without opening anything.
- Auto-scaled sizes — output tiles show file sizes in B, KB, MB, GB, or TB depending on the file, so the number you see always matches the scale of the file.
See Assets Catalog for the full library walkthrough.
Selected project assets appear as context chips in the Video agent composer. Remove a chip if you do not want that asset used in the next request.
Video Plugins
Video plugins package repeatable video workflows. They can prepare research, pick B-roll, generate images or clips, choose music, assemble the timeline, run checks, and render a preview depending on what the plugin was built to do.
To run a video plugin:
- Open the Video agent dock.
- Click the plugin picker.
- Search or browse the available plugins.
- Review any capability prompt that appears.
- Click the plugin to apply it to the current project.
- Send the prepared prompt to the agent.
Some plugins need review before they can use network, media generation, vision, transcription, analysis, video editing, or YouTube capabilities. If a plugin was imported or changed since your last review, VideoMode asks you to review it again before those capabilities are allowed.
After a successful plugin-backed render, VideoMode may suggest saving the flow as a reusable plugin. You can dismiss the suggestion, save it to the current project, or save it to your user plugins so it is available in other projects.
Editor Handoff
Use Editor handoff from the Preview toolbar when you want to finish the cut in a desktop editor.
- Choose target formats: Final Cut Pro, Premiere Pro, DaVinci Resolve, OTIO, EDL, or CapCut fallback.
- Choose whether to Copy media into the package or Link media from its current location.
- Export the handoff package.
- Open the package folder or copy the package path.
The package includes timeline interchange files, captions, cut list, media records, a conformance report, action history, checksums, and a reference render when one is available. Your audio edits — clip gain, fades, track volume, crossfades, and ducking — travel with the package, along with provenance for any generated audio. Some targets are fallback or compatibility formats, so review the conformance warnings, including any note about manually verifying audio edits, before sending the package to another editor.
Sharing
Open Share from the project toolbar to package a finished render with its captions, transcript, and provenance metadata. Shared bundles never include provider keys.
Rendered exports can also include a credits sidecar for required attribution, embedded metadata, and best-effort C2PA signing when signing is enabled for the workspace.
Tips
- Save a template when a project layout works — future projects can start from the same scaffold.
- Generate a proxy before final render to preview transitions and timing quickly.
- Run Auto color and Loudness before publishing to keep cross-device playback consistent.
- Use Linked Sources to keep large source files outside the project folder while still referencing them.
- Check the QA report before shipping — silent gaps and missing media are easy to miss in preview.
Related Docs
- Assets Catalog — Browse, search, and attach items from your unified library
- Media Generation — Image and video provider setup
- Voice and Speech — TTS provider configuration
- DesignMode — The matching design and document workspace
- Workspace Security — How local workspace access is protected