ToolMyth Video Studio

About ToolMyth Video Studio

ToolMyth Video Studio Beta edits and exports video locally in your browser. Arrange local video, music, images and text, with manual captions, SRT/WebVTT import, voiceover recording, picture-in-picture, green screen and masks, alongside crop, speed, freeze frames, transitions, filters and animation. Use multi-selection, grouping, ripple edits, markers, local autosave and project files for longer edits. Export MP4 or WebM where supported, with 480p, 720p and up to 1080p presets and a mixed audio track.

The editor keeps source media separate from your decisions. Moving a clip changes when it plays; trimming changes its range. Cropping and transforming change the composition without altering the original file. Text remains editable text rather than a flattened picture.

Current capabilities

Finishing Tools extends the same timeline and export engine. Media-free adjustment layers process composited lower layers during their visible interval; higher overlays and captions stay above them. Color/filter settings, opacity and bounded color keyframes are supported. Import local SDR .cube LUTs (one 1D table up to 4096 entries or one 3D table up to size 33, 4 MiB file limit), blend strength, edit Master/R/G/B curves and save local effect presets with explicit Replace/Append/Cancel.

Compounds store nested sequence references without flattening sources. Open a compound with its command or desktop double click; use the breadcrumb to return to Main Timeline. Up to 12 additional sequences and five nesting levels share the media library. Duration changes reject collisions unless Ripple can safely shift following clips. Duplicate landscape, vertical or square sequences; existing clips are fitted rather than automatically reframed.

Optional Color Scopes show an RGB/luma histogram, luma waveform and RGB parade from reduced preview frames. Scopes start off and update at five Hz, or two Hz in Performance Mode. These lightweight SDR analysis tools never affect output and are hidden on small mobile screens.

Clip volume automation combines with normal fades and equal-gain/equal-power crossfades on audio tracks. Numeric point editing is in Finishing Tools; optional desktop timeline envelopes support direct points. Track insert EQ/gate/compressor/limiter and basic rumble/hiss filtering feed a master gain/effect chain. The same processing graph is used for offline export. Track/master dynamics process continuous sections limited to 120 seconds desktop or 30 seconds mobile.

Render Preview In/Out creates an explicit silent 480p editing copy while live project audio continues. Caches are limited to three entries and 64 MiB, with 30-second desktop or ten-second mobile ranges. Project, sequence, LUT, audio and source changes invalidate cached playback immediately; final export always uses originals. Clear current-range/project/all Video Studio caches without clearing unrelated data.

Export Now or add fixed snapshots to Render Queue, with up to five pending/eight total jobs. One render runs at a time while the page stays open. Failure/cancellation stops the queue unless Continue is enabled. Waiting metadata can restore locally; interrupted renders become failed and matching source files must be relinked. Download completed outputs manually; only two recent blobs within 96 MiB desktop/32 MiB mobile are retained. Export presets can be saved locally.

Generate Captions in the Captions panel processes a selected video/audio/voiceover clip locally. First use downloads about 65 MB of runtime/model data; model files are cached where possible, up to 96 MiB. Speech language offers Auto Detect plus a small supported selection. Fast and Balanced use the same quantized multilingual Whisper Tiny model. Measured word timestamps drive color/background/pop/underline highlights in both preview and export; caption-level timing is used when word timing is not requested or alignment is uncertain. Long timed creator captions page at measured word boundaries to honor words-per-line and maximum-line settings.

Creator Tools includes common 16:9, 9:16 and square templates, a fact-video layout and simple presentation. Titles, captions, intro/outro and CTA elements remain ordinary editable clips with normal animation keys. Save/export your own structure/style templates without source media, then replace media placeholders from imported originals. Up to 10 local templates share a 4 MiB budget. Templates are not uploaded.

Track Motion explicitly analyzes a cropped source region at reduced resolution and 3-5 frames per second. Draw a region or select a detected face using the local detector. Apply tracked blur/pixelation through the existing mask compositor or attach an editable text/image path using normal keys. Stop/correct lost targets; blur/pixelation covers the full source after the marked lost interval. Tracking is not a privacy/redaction guarantee. Manual crop remains available; AI reframing and automatic scene cuts are deferred.

Audio cleanup provides three-band EQ, conservative Voice/Music presets, a Voice Enhance preset, compressor, approximate limiter, RMS noise gate and basic rumble/hiss filters. Preview uses live Web Audio nodes and export reconstructs the same chain continuously per processed clip. Reset Audio Effects retains timing and volume. Detect Silence analyzes downsampled audio, previews detected ranges and requires explicit selected-range confirmation before one-undo ripple deletion.

File offers New Project, Open Project, Save Project File, Locate All Missing Media and Project Storage. Lightweight versioned .tmvideo project files hold timeline, effects, captions, tracks, groups, links, markers, ranges and opaque local media IDs; they do not contain source files, object URLs, file handles or local paths. Invalid or unsupported files are rejected before the current edit is replaced. Local IndexedDB autosave runs about 1.5 seconds after meaningful edits, keeping a current and previous metadata snapshot. Recovery offers a choice on reopening, with real saved timestamps; it never silently replaces your edit.

Supported images up to 25 MiB, audio up to 64 MiB and video up to 200 MiB are automatically cached in local IndexedDB, subject to a 500 MiB total cap and available-storage safety margins. File System Access handles are also kept where supported; permission may need to be granted again. Large files without a usable handle remain session-only and need relinking after reopening. Media stays on this device and is never uploaded to ToolMyth. Recovery first tries cached media and already-permitted handles; missing sources remain visible as Media unavailable with Relink and Locate All Missing Media actions. Multi-file matching uses name, size, duration and dimensions; mismatches require confirmation and incompatible types/source ranges are rejected. Final export blocks required offline originals. Voiceovers and freeze frames use the same bounded local cache and can be downloaded individually.

Select mode supports Shift/Ctrl/Cmd clicking and desktop drag selection in empty track space. Selected groups move together with relative timing preserved; copy/paste duplicates clip metadata while reusing source media. Video/audio links are separate from groups. Linked Selection normally selects and moves/trims a linked pair together; disable it or Unlink for independent work. Group, link, split and multi-delete commit as one undo transaction. Locked tracks reject edit commands without stopping playback.

Ripple Editing is off by default. When enabled, a move shifts the selected downstream block on participating tracks, and trims shift following material. Normal Delete always leaves a gap; explicit Ripple Delete closes selected intervals, and Ripple Delete Range cuts/closes the In/Out interval on selected tracks or all unlocked tracks. Unsafe collisions, locked peers or desynchronized group/link moves are rejected atomically. Insert at playhead splits material and makes room on the target and linked tracks; Overwrite trims/replaces that interval without shifting later content. Append places media after the final clip on the destination track.

Timeline snapping uses an eight-pixel tolerance at the current zoom and includes clip/caption edges, markers, transition boundaries, range edges and the playhead. A guide shows the active snap. Mark In/Out with I/O, play or loop the range, and choose Entire Timeline or In/Out Range in Export. Range export keeps absolute source, fade, transition and keyframe timing while writing output from zero. Timeline markers can be renamed, retimed, deleted and navigated; they remain at project time during ripple edits.

Timeline menus manage one base plus one PIP video track, up to four audio tracks, four image/text overlay tracks and two caption tracks. Rename, lock/hide or mute/solo, collapse, and reorder same-kind tracks; only empty non-base tracks can be removed. Overlay/caption ordering affects composition. Shift-wheel scrolls horizontally, zoom keeps the playhead position anchored, and Fit Timeline shows the project. Offscreen clip details are culled; the playhead updates separately from the track tree.

Generate Proxy explicitly encodes a local 480p preview copy through the existing native export engine. It is cancellable, keeps source duration and audio, and uses a session-only cache capped at 128 MiB total / 64 MiB per proxy. Source/aspect/codec and export-memory limits apply. Proxies are never generated automatically, and final export always reads original Blobs. Delete a proxy or Clear Proxies without touching originals. Proxy controls are hidden on phones; a larger screen is recommended for advanced workflows.

Keyboard Shortcuts in View lists the central command registry. Space plays/pauses; S splits; Delete leaves a gap; Shift+Delete ripple-deletes; N toggles snapping; M adds a marker; I/O sets the range; Home/End and arrows navigate. Ctrl/Cmd handles undo, copy/paste, duplicate, group and project-file save. Timeline-focused Ctrl/Cmd+A selects clips; text inputs, captions, numeric fields and dialogs keep their own keys. Alt+M mutes a clip and Alt+arrows nudge visuals. Shortcut remapping and clip-local markers are deferred.

Two dedicated caption tracks hold editable caption text, separate from text overlays. Add Caption at the playhead, or import local UTF-8 SRT/WebVTT files (up to 1 MiB / 1000 total cues). Valid cues survive malformed blocks; warnings identify skipped ranges and retained overlaps. Basic text markup is stripped; cue positioning, styles, karaoke and advanced subtitle features are ignored. Search, edit start/end, move, trim, split, duplicate, change tracks or apply a global timing offset. Export SRT downloads the selected caption track. Generate Captions locally uses an optional Whisper Tiny model to create ordinary editable caption clips with measured word timing where returned. Word highlighting, creator styles, merge and text-cursor splitting share the normal caption system. Editing text clears stale word timestamps.

Caption styling includes four presets, safe fonts, size/bold, alignment, text/background colors, background opacity, top/center/bottom placement, width and line spacing. Long lines wrap inside a 5% title-safe region. Captions draw above the visual layers using the same text renderer at full export resolution. Overlaps on a track are retained and warned about, with separate timeline lanes for selection.

Record Voiceover requests microphone permission only after Start recording. Your microphone is used only for this recording session. Optional 3-2-1 countdown starts after permission. Capture begins at the playhead while the timeline plays, pauses during buffering, and has a timer and Stop control. Mute timeline during recording affects monitoring only. Microphone input is never routed to speakers. The local recording becomes an ordinary audio clip with waveform, trim/split, volume, fades and mute. Microphone tracks stop after completion, cancellation, failure or leaving the editor. Recording is limited to five minutes / 64 MiB and requires a secure context and browser MediaRecorder support.

Add as PIP creates a picture-in-picture video or image on one dedicated PIP lane. Five position presets include margins; use normal visual transforms and keyframes afterward. Overlay video audio starts muted and can be enabled in Audio controls. One base plus one PIP video layer is supported; base transitions temporarily need an additional handle decoder. PIP appears below Overlay 1 and the caption tracks.

Green Screen / Chroma Key offers green, blue or custom key color, similarity, edge softness and spill reduction. Rectangle/ellipse masks support local position, size, rotation, feather and inversion; Blur inside mask uses a bounded shader blur. Desktop preview mask handles move, resize and rotate; numeric controls work on mobile. Masks use clip-local project units measured from the fitted source center and follow transforms/keyframes. Actual effect order is source, crop, chroma, mask/region blur, color/filter, transform, opacity, layer composition, then captions. All effects appear in local MP4/WebM exports. Keyed transparency reveals lower video/image layers or the project background; ordinary MP4/WebM exports flatten the final composition.

Export uses native WebCodecs encoding and a separate full-output-resolution canvas, sharing the preview compositor and timeline interpretation. MP4 uses H.264/AAC where supported; WebM uses VP9 or VP8 with Opus. The dialog probes the current dimensions, FPS and codecs rather than assuming browser support. It offers only compatible formats and never silently drops requested audio. Unsupported MP4 configurations can use WebM; browsers lacking the required native APIs need another compatible browser. No FFmpeg compatibility runtime is included in this exporter.

Project/480p/720p/1080p presets preserve orientation and aspect ratio. Custom even dimensions stay within a Full HD area and a 1920-pixel edge; a different aspect ratio fits the composition over its project background. FPS choices are 24, 25, 30, 50 and 60. Low/Standard/High quality chooses a default bitrate, with an optional manual bitrate. Fast Draft uses 480p class, 30 FPS and lower quality. File size is an estimate rather than a guarantee.

Audio export mixes source/video audio, detached references and music internally in short 48 kHz stereo blocks. Clip gain, mute/solo, visibility and fades follow the editor rules. Preview master mute/volume affect monitoring only and do not change the exported mix. Video speed changes preserve source-time alignment, but offline resampling can change audio pitch; the dialog warns before rendering. Audio Off is an explicit setting. Export does not record speakers or re-record the preview; recorded voiceover is mixed from its local audio asset.

Render progress counts actual frames and stages. Cancel stops the export, releases temporary encoder/decoder resources and returns to the same project. Success is shown only after output dimensions, duration, tracks and sample decoding pass validation. Download is manual; only the current output stays in memory, and it is released when replaced or the dialog closes. Export does not save a reusable project file.

Constant video speed ranges from 0.1x to 8x, with common presets. Source in/out remain unchanged while timeline duration changes. Later clips stay put; speed changes that overlap another video or exceed project limits are rejected. Native pitch preservation is requested, but quality and audible speed ranges depend on the browser. Unlink detached audio before changing video speed; detached audio keeps its independent timing. Restore 1x before detaching new audio.

Freeze Frame captures the paused decoded source frame locally, up to a 1920-pixel edge, and adds an image overlay at the playhead for a configurable duration (three seconds by default). It copies the current visual appearance as editable transform/crop/effect metadata. It does not stop or shift the underlying video or audio. Undo/redo also restores the generated image resource.

Adjacent clips on the same video track support Cut, Cross Dissolve, Fade to/from Black, Dip to Black, Slide Left/Right, Zoom and Wipe Left/Right. Transition controls belong to the incoming clip. Blend/movement transitions use spare source frames around the cut; trim source in/out to leave handles, then make the clips adjacent. Durations default to 0.5 seconds and are clamped to the available media within 0.1–3 seconds. Black fades/dips can use visible clip edges. Timeline indicators open transition settings. Visual transitions keep audio on its existing hard cut; they do not add audio fades.

Each visual clip supports up to five ordered effects, including basic color correction. Filter looks include Grayscale, Sepia, Warm, Cool, Vintage, High Contrast, Fade and Black & White. Brightness, contrast, saturation, exposure and approximate temperature/tint use native Canvas filters at preview resolution. Toggle, reorder or delete effects and use Reset Color for neutral values. These are creative adjustments, not professional color grading.

Video, image and text clips support clip-local Position X/Y, Scale X/Y, Rotation and Opacity keys with linear, ease-in, ease-out and ease-in-out interpolation. Add/update/delete keys at a paused playhead, navigate previous/next keys and click selected-clip timeline markers. Fade, slide and zoom presets generate the same keyframes; replacing an existing animation is explicit. Split/trim rebase visible keys and sample boundary values; easing inside shortened segments can change slightly. Reset Visual Adjustments clears transform, crop, color/effects, keying, masks and animation while keeping source timing, speed and audio.

Up to four dedicated audio tracks support music, dialogue and detached video audio. Multiple audio clips can mix during preview, with source-offset trimming, split, move, copy/paste, volume, clip/track mute, multi-track solo and linear fade in/out. A master volume/mute and activity meter affect monitoring without changing clip levels.

Detach Audio creates a source-linked clip with the same timeline position and source range, and mutes the original video audio to avoid doubling. It references the existing file rather than extracting a new one. Linked clips normally select and move/trim together through Linked Selection; disabling that toggle or using Unlink permits independent timing. No audio file is downloaded or exported.

Waveforms use a cached, downsampled amplitude envelope. Small sources up to 8 MiB and two minutes can be decoded temporarily where supported. Long sources stream normally without a full waveform decode; video-audio waveforms are omitted when the browser cannot decode them directly. Waveforms describe the source, while an envelope line shows the clip fades.

Timeline editing includes multiple video clips, trim, split, move, delete, snapping, zoom, copy/paste and duplicate. Overlay 1 holds overlapping images and text, each with editable timing. Order commands determine which overlay is drawn above another. Track visibility and locking help you inspect or protect a layer without deleting it.

The canvas compositor draws the project background, one base video, the PIP layer, image/text overlays and caption tracks in their stored layer order. It supports non-destructive crop, resize/transform, rotation, horizontal and vertical flip, opacity, transparent PNG images and multiline text. Browser-safe fonts, text backgrounds, title/caption presets and optional approximate safe-area guides are included.

Positions use project coordinates with the origin at the top-left and each transform centered on its object. Changing canvas dimensions preserves proportional positions and keeps scale, rotation and crop decisions. Fit shows the whole cropped source; Fill covers the project frame and can hide edges. A smaller preview buffer keeps rendering responsive without allocating a full output-resolution canvas.

Current limitations · Beta

• Local speech recognition can be slow and inaccurate, particularly for music, accents or quiet audio. First use requires a model download and roughly 0.5-1 GB free memory. Transcribe up to 10 minutes per desktop section or two minutes on mobile/low-memory devices, using In/Out for longer sources. Cancel terminates the speech worker. No transcript or media is uploaded; model hosting receives ordinary model-download requests.
• Tracking is limited to forward sections up to 60 seconds on desktop or 20 seconds on mobile, with at most 601 stored points. Low texture, occlusion and fast motion can lose the target and need manual correction. General object recognition, backward tracking, automatic reframing and semantic object detection are deferred. Review all sensitive frames.
• Audio cleanup export is limited to 120 seconds per clip on desktop or 30 on mobile, with a 48/16 MiB processed-audio cache. This keeps gate/compressor state continuous across export blocks. Split longer clips or reset effects before export. Noise reduction is basic filtering, not neural denoising or voice isolation. Multiple loud clips can still overload the mix; the meter indicates clipping.
• Chroma key is basic, not studio-grade. Green screen, masks and region blur require WebGL; unsupported devices show an explanation and cannot enable them. AI background removal, Bezier masks and PIP border/shadow/corner styling are deferred. Creator Tools provides bounded forward region tracking and tracked blur/pixelation. Local face detection uses native FaceDetector where available, otherwise an optional 1.27 MB UltraFace model through the shared CPU/WASM runtime. Choose among detected faces, or draw a region manually. Automatic semantic object/license-plate detection is not included.
• Two video layers increase CPU/GPU and decoder memory use. Use 720p/30 FPS on weaker devices. Mobile PIP is limited to video sources up to 720p; export enforces this too. Performance mode reduces the shared preview to 360p including PIP, keying and masks without changing final project/export settings. Blur is a lightweight fixed shader approximation, not a guarantee that sensitive details are irrecoverable. Microphone permission/format support varies by browser; voice recordings remain in memory unless you explicitly enable generated-media caching or download them.
• Export is limited to up to 1080p initially; 4K export, lossless/broadcast mastering and unlimited export duration are not supported. Desktop can downscale browser-decodable sources up to 4K; mobile source decoding is limited to Full HD. Reverse, speed ramping, motion blur, HDR grading and pitch-preserving offline time stretching remain deferred. Bounded SDR LUTs, curves and adjustment layers are available in Finishing Tools.
• Export speed and memory use depend on device, source codecs and complexity. The dialog enforces both duration and estimated-output limits: desktop ceilings are 10 minutes at 1080p (five at high FPS), 20 at 720p and 30 at 480p, with lower effective limits when the output budget is reached. Current output is capped at 256 MiB on desktop (128 MiB with low-memory hints) and 64 MiB on mobile. Mobile is limited to 720p class and three minutes, or five minutes at 480p, also subject to its output cap. Try lower FPS, lower quality or a shorter project for demanding edits. These caps do not guarantee every device can finish every project.
• One base video plus one PIP video can play alongside image/text overlays, two caption tracks and up to four audio tracks. The initial limits are 20 combined video/audio assets, 50 images, 10 simultaneous visual overlays, 10 text clips, eight simultaneous clips across the audio tracks and a 30-minute timeline. Additional PIP lanes are deferred. Projects are capped at 1500 total clips, including up to 1000 captions, and 200 markers. Captions are limited to 1000 total per project.
• Browser audio codec support varies for MP3, WAV, M4A/AAC, OGG and WebM. Large projects may use significant memory; large/long waveforms are deliberately omitted. Playback may briefly buffer on source changes or seeks. Gain above 100% and overlapping loud sources may clip; monitoring does not automatically normalize or master the mix.
• Performance depends on your browser and device. Codec support depends on the browser; an MP4 or MOV extension does not guarantee decoding. Video seeks, source switches and two-decoder transitions can buffer briefly. Canvas filter support is required for color/effects; unsupported browsers show a message. Preview uses Auto/720p/480p bounds; optional Performance mode reduces the preview to 360p without changing the project. Frame stepping follows project FPS but is not a guarantee of frame-perfect decoding. Backgrounding the page pauses playback.
• Keyframes cover transforms, opacity, text reveal, adjustment color and clip volume, with up to 60 keys per channel. Detailed curves, scopes, audio envelopes and nested editing are intended for desktop/tablet. Mobile exposes simpler preset/LUT/curve reset and queue controls. Pan, vectorscope, HDR/ACES grading and professional true-peak metering are not implemented.
• PNG, JPEG and WebP are supported as static image overlays. Animated WebP uses a prepared static frame; GIF animation and SVG import are not supported. Images above 50 MiB, 64 megapixels or 32768 pixels per edge are rejected for browser safety. Accepted original images remain in the session, with bounded previews.
• Crop uses inspector percentages and Free/common aspect presets. Preview corner handles preserve proportions; independent X/Y scale is available numerically. Safe guides are approximate and do not guarantee permanent platform-interface dimensions.
• Autosave is local recovery assistance, not guaranteed backup or cloud sync. Browser storage limits, private browsing, clearing and eviction can remove saved metadata or opt-in generated assets. Keep downloaded project files and source/generated media separately. Project files do not pack media, and original files may need relinking after refresh. Proxies are session-only and must be regenerated.

What happens to your media?

Editing and export run locally in your browser. No rendering upload is required. Media, audio, frames, waveforms, filenames, overlay text, effects, keys, timeline data and output video are not sent to ToolMyth servers, Cloudflare storage, analytics or external processing APIs. Large projects can require significant CPU and memory. Exported files are downloaded directly from a local browser Blob; final output is not persisted in browser storage; explicitly requested preview renders are bounded temporary IndexedDB blobs. Project metadata, templates and optional generated-asset copies may be stored only in local browser IndexedDB; optional speech/face-model files use a dedicated local model cache. Model data downloads from Hugging Face only when caption generation or local face detection is requested; no media or transcript is sent with these requests. Project Storage can disable autosave and clear Video Studio snapshots/generated cache; proxy copies are session-only. This is not remote storage, cross-device sync or online backup.