Changelog

One version instead of a choice#

AI montage now delivers one version of the reel. The director used to build an alternative with a different script as well, but it came out weaker than the main one — that time now goes into a self-check. There is nothing to pick: when the montage is ready, the project opens on the Editor section with the finished reel. The Pick a version sheet and the button with the number of versions in the project header no longer appear after a montage.

A rebuild with new settings and a retry work the same way. The montage price hasn't changed — 24 tokens.

The director checks the finished reel itself#

Once the reel is built, the director reviews it the way a viewer will: in playback order, with what is on screen and what can be heard. It straightens out the logic and order of the clips and removes cut-off thoughts, and if the plan changed, the reel is built again. On the run card this is the Polishing the cut stage; in the Chat journal — Director’s reasoning on the finished cut.

A faster montage#

Slicer watches the frames while it transcribes the speech, and prepares the editor preview while the director is thinking about the edit. The finished reel arrives noticeably sooner.

Earlier versions stay in the history#

Every rebuild, retry, style change and export still adds a new version, and the earlier ones stay in Media → History: you can bring back any of the last 10 with Make current or by asking the director in the Chat — “bring back the previous version”.

API and MCP#

• AI montage delivers one version: versionChoice is null after a montage, and the project.version_ready event doesn't arrive.
• POST /v1/projects/{id}/versions/{versionId}/choose and the MCP choose_version tool now also make current any version from the history that has a file (hasFile) — for example, the reel from before a rebuild.

The self-check of the finished reel · The finished reel · Video versions in the API

A new Sound first transition#

This is how professional editors cut speech: the next clip's sound starts a split second earlier, the previous picture holds a moment longer and changes on the first words of the new sentence. The cut never lands in a pause, the speech flows evenly, and the picture change is barely noticeable.

Sound first is the first tile in the Transition panel and in the Library. Its duration is how far the picture lags behind the sound: 0.3 s, 0.5 s (default), 0.65 s or 0.8 s. The cut shows up in the preview right away — on the web and in the app.

AI editing adds it on its own#

At speech cuts where the picture visibly changes — another take, another source — Slicer adds Sound first on its own and picks the lag by the first word of the new clip. Where an abrupt cut is the point, the director keeps a straight cut. You can remove or change any of these cuts like a regular transition.

API and MCP#

The tune_block and set_all_transitions operations accept a new jcut transition kind with a duration from 0.15 to 1 s.

Transitions between clips

The single edit operation catalog has grown — the new operations work in POST /v1/projects/{id}/edits, in the MCP tool edit_project, in the AI assistant and for the director in the Chat. All of them are free.

Cutting and fragment look#

• split_block, trim_block, duplicate_block — split a fragment at a video moment, trim or extend it at the start or end, duplicate it.
• freeze_frame and set_hold — a freeze frame and its length.
• replace_source — show another source in a fragment; in a fragment with speech the new picture goes over it as an insert.
• set_process — free processing: reverse, stabilization, noise reduction, audio enhancement.
• apply_keyframe_preset — a camera move across the whole fragment: zoom in, zoom out, pan.
• tune_block — the new background field: a blurred background or a colour behind a fitted picture.

Music and voice#

• set_music, remove_music — a track from the sources or project media, volume, ducking under speech, start, fade-in and fade-out.
• fit_music, fit_to_beats — the track fitted to the video length, footage cuts snapped to the beats.
• add_voice, move_voice, trim_voice, set_voice_volume, remove_voice — a voice recording on the voice track.

Inserts and elements#

• add_insert, update_insert, remove_insert — your own photos and videos over the video, full frame or in a window.
• update_plaque, update_card, update_zoom — change a plaque, a card (label, your own photo) and a zoom in place.
• set_element_position — move an emoji, card or plaque around the frame.
• set_phrase_text, reset_phrase_text — the caption text of a whole phrase.
• move_element — now also moves sounds and markers.

Captions from speech#

• POST /v1/projects/{id}/transcribe — recognise the speech in the sources, like Captions from speech in the editor: after that a project without the AI director gets captions, and words with ids appear in the video text. The sources and language fields are optional. Priced at cost per minute of sound: without confirmCost the response is the price (a Quote with estimate and cost), with it — a start in the background.
• GET /v1/projects/{id}/transcribe — recognition jobs: processing, ready or failed (then the tokens come back).
• POST /v1/estimate — the new transcribe kind (needs projectId).
• MCP — the transcribe_speech and get_transcribe_jobs tools.

New skip reasons: block_or_source_not_found, audio_not_found, no_music_or_beats, recording_not_found, media_or_place_not_found, word_not_found. Clients don't need to change anything — all changes are within /v1.

Editing the video · Captions from speech · Music and voice · Your photos and videos over the video · API version history

Captions got livelier: the spoken word can pop, sit on a box or get an underline, a phrase can appear word by word, and with the Auto position each phrase finds free space in the frame by itself.

New designs#

• Pop — 2–3 words on screen: the spoken word pops out in the accent color while the others make room.
• Box — a line of up to six words: the spoken word sits on an accent box that slides from word to word.
• One word — one big word at a time, in sync with speech.

The Pop template on the wizard's “Style” step pairs “Pop” with the Punch font and the Lime accent — for energetic videos.

Caption options#

The Caption style sheet, the project's Settings section and the wizard's Captions step have four new options that work with any design:

• Word highlight — Color, Pop, Box or Underline: for example, in “Karaoke” the spoken word can now grow;
• Entrance — Fade, Pop, Slide up or Word by word;
• Words on screen — One, 2–3 words, Line or Phrase;
• Letter case — CAPITALS or As spoken.

All of them default to As in the style. Picking a design resets the highlight, entrance and words on screen to it, so a new design shows up the way it is meant to. Templates keep the new options too, and the director in the Chat changes them on request: “Make the spoken word pop”.

Smart Auto#

With the Auto position, each phrase goes where the frame has the most free space: off the face, the hook, plaques, titles and picture-in-picture windows. A phrase right after another keeps its place, and captions return to their usual place on a cut — the text doesn't jump without a reason. Emoji and media cards go next to the captions, wherever they are at that moment. Slicer knows where faces are after an AI montage or Follow the face, and Auto always steers clear of the reel's elements.

Public API 1.8.0 and MCP#

• subtitles in the style (PATCH /v1/projects/{id}/style, the MCP tool set_style) and in montage settings has the fields highlight, animation, words, letterCase and speakerColors, and the designs pop, box and word; the pop template.
• The GET /v1/styles catalog has subtitleOptions: the caption options with names and values in both languages.
• A partial subtitles change now keeps the other fields — only what you pass in the request changes. A new design without highlight, animation and words resets them to auto.

Clients don't need to change anything — all changes are within /v1.

Designs · Caption options · Smart Auto · Style in the API · API version history

Edits in words — almost everything the editor does#

Ask the director in the Chat, and it does it for you:

• cutting — split, trim, reorder and duplicate fragments, add a freeze frame, replace a fragment's picture with another source, cut pauses;
• music — set a track from the sources or project media, volume, ducking under speech, fit it to the reel length, snap cuts to the beats;
• the voice track and your own photos and videos over the reel — full frame or in a window;
• elements — edit callouts, cards and zooms, element positions on the frame, the caption text of a phrase;
• free fragment processing — reverse, stabilization, noise reduction, audio enhancement — and preset camera moves.

For example: "Use the track from the sources, make it quieter and fit it to the reel length" or "Show the photo of the finished cake in a window over the phrase about the result".

Style without a re-cut#

The director changes the template, font, colour, look, size and position of the captions, speaker colours, sound accents and the intro zoom — the new style shows up in the preview right away.

Everything the director changed in one reply is a single undo step: Undo in the editor brings the reel and its style back. Paid AI operations and export come as an Open in the editor button: you see and confirm the price in the tool itself. You can rename the project right in the Chat.

Chat with the director · Reel style · What is still done in the editor

Report a problem#

Report a problem is in the project's “⋯” menu, on the card of a failed run, at the bottom of Help and in Profile. Choose what's wrong, describe it, attach screenshots and, if you like, project data for diagnosis — the reply comes by email and shows up in My reports.

Rating the app#

After a successful export Slicer sometimes asks How do you like Slicer? — stars and an optional review. In the iPhone and iPad app, iOS may then offer an App Store rating. You can also rate from Profile → Rate Slicer.

Usage statistics#

Slicer counts which screens and features people use — no text, files or IP address, no third-party analytics services. To turn it off, use Help improve Slicer in Profile.

Details — Ratings and problem reports and Data storage.

Over the Slicer MCP server, your AI assistant now edits the video — just like you do in the app with Edit it myself. Connect any client that supports MCP, give it your materials, and it builds the video with editor commands: cuts, hook, titles, plaques, zooms, emoji, transitions, colour, music, caption style. All of it is free.

What changed#

• create_project creates a project without the AI director by default (mode: "manual"): free, with no price and no confirmation; the materials go onto the timeline in order.
• Reading the video (get_transcript), edits with the whole operation catalog (edit_project), style (set_style), auto colour (auto_color), export and publishing (export_video) are free.
• New tools transcribe_speech and get_transcribe_jobs — captions from speech for such a project: at cost per minute of sound, in the background.
• The make_reel prompt follows the free path: a project without the AI director → captions from speech with the price, if the materials contain speech → edits → export.

What's paid#

Only Slicer's own AI features spend tokens, and each one names its price first — the assistant starts it only after you agree:

• an AI montage by the Slicer director — create_project with mode: "ai", remount_project, retry_project; the assistant offers it only if you ask for it;
• captions from speech — transcribe_speech;
• the Slicer AI assistant — ask_assistant;
• caption translation — translate_captions.

If your client called create_project without mode and expected a montage price, pass mode: "ai" explicitly. The REST API hasn't changed: the default mode of POST /v1/projects is still ai.

MCP server · What's free and what's paid · Captions from speech · Example scenarios

Your AI assistant can now edit a video by itself — on its own subscription, and for free in Slicer. The assistant looks at the frames and writes the plan, and Slicer applies the rules of its AI director to the plan on our server: whatever it can, it corrects itself, and for the rest it answers with issues. The rules themselves are not passed to the assistant.

What's new#

• Director brief — get_director_brief: the material — speech by word ids, pauses, retakes and your wishes — a few general tips and the schemas of the plan and of observations from the frames. The video type, pace, effects and length can be passed right away — they are saved to the project.
• Frames — get_frames: the assistant looks at the sources itself, up to 16 frames per call, at the seconds it needs or evenly.
• Accepting the plan — submit_director_plan: the plan and the observations from the frames become the video — a new revision. The answer has what Slicer corrected itself, issues with the plan and the video as text; the assistant fixes the plan accordingly and submits it again.
• The direct_reel prompt — the whole workflow in one command.
• API 1.7.0 — the same routes: POST /v1/projects/{id}/director/brief, GET /v1/projects/{id}/sources/{index}/frames and POST /v1/projects/{id}/director/plan.
• The director's rules stay on the server. In the Chat, the The director’s brief entry is now hidden by default.

What's paid#

The brief, frames, accepting the plan, edits and export are free. Only speech recognition for captions (transcribe_speech) is paid — at cost per minute of sound and after you agree to the price. An AI montage by the Slicer director stays a paid option for those who don't have an assistant of their own.

Your own director · Tools · The direct_reel prompt · The same in the API

Two cuts for one price#

A montage, rebuild or retry now makes two cuts of the reel: the main one — following your settings and notes — and an alternative with a different script, pacing and effects. When the montage is ready, the Pick a version sheet opens: watch both and pick the best. The montage price hasn't changed — 24 tokens.

Rarely, with very large footage, the alternative may not work out — then you get one cut.

Pick a version from the Chat at any time#

Ask the director in the Chat — “bring back the second version”, “keep the main one” — and it makes any version of the latest montage current, even after the version choice is closed. The reel on the timeline is replaced with the chosen version, and the previous cut stays in Media → History — you can bring it back with Make current.

Pick one of two cuts · Chat with the director · How the director works

A media card by hand#

The Elements mode has a new Card button in the Add at the playhead block. The Media card sheet has two tabs:

• Own photo — a project photo or a new one (Upload a photo, in the app — From Photos) and a caption. Free: the photo goes into the card with rounded corners, a frame and a shadow, just like in the final video, and the preview shows it right away.
• From a description — describe the image in the What to show field, and the AI draws it in 10–30 seconds. The price is the cost of one image (≈0.65 tokens), shown on the button before you start; tokens are charged only if the image is made. The image is saved to the project media — you can use it in other cards too.

A card in the Media cards list and a selected card on the strip have Replace photo: another photo or a new image, with the timing and caption kept. The project history has a new kind of work, Card from a description.

Export for an editing app#

The project's “⋯” menu → Export for an editing app: finish a Slicer edit in a professional app. The formats are Final Cut Pro (FCPXML) for Final Cut Pro and DaVinci Resolve, OpenTimelineIO (OTIO) for DaVinci Resolve, Premiere and Avid, and EDL (CMX 3600) for any app.

The archive holds the timeline, SRT captions and a README. What's transferred: clip order, source in and out points, speed and freeze frames, transitions that overlap frames (as cross dissolves), music on its own track, markers and titles (in FCPXML). Color, effects, masks, 9:16 reframing and the reel's graphics are not transferred — you redo them with the app's own tools. The reel's source files download one by one in the same sheet. Free.

A media card by hand · Export for an editing app · Operation prices

Frame#

• Follow the face — a new button on the Frame editing layer: the crop window smoothly follows the speaker through the whole clip, with no jerks or jitter. If the AI montage has already found the face, following turns on right away and for free. Otherwise your iPhone, iPad or Mac finds the face itself — for free — or an AI in the cloud does it at cost. You can adjust a following frame by dragging at any moment of the clip.
• Chroma key in the Background panel: a green or blue screen turns transparent, with a color or a project photo behind. The screen color is picked by itself from the frame edges or with an eyedropper on the preview; Similarity, Smoothness and Remove fringe polish the edge. No AI, and free.
• The mask now moves across the frame, rotates and inverts, and Animation point lets it travel along with an object. Drag the shape right on the preview; in the app, twist with two fingers to rotate it.

Motion#

• Keyframes do Rotation and Opacity, and each point has an easing to the next one: Smooth, Linear, Ease in or Ease out.
• Volume points: Volume point in the Volume panel — a clip's sound rises or fades smoothly between points.

Speed#

• Speed curves — the Curve tab in the Speed panel: six ready-made curves — Montage, Hero, Bullet, Flash, Ramp up, Ramp down — and your own on a graph with points. For b-roll without speech.
• Smooth slow motion creates frames between the real ones: a slowed-down clip of up to 10 seconds moves smoothly, as if shot at a high frame rate. Free, made in the background.

Public API 1.5.0#

New edit operations: set_framing (with track — face following), set_keyframe, remove_keyframe, set_mask, set_chromakey, set_speed_curve, set_smooth, set_volume_keys. They're also in the MCP tool edit_project.

Follow the face · Chroma key · Keyframes · Speed curves · API operations

Fit to length#

Fit to length is a new button in the music's Volume panel. The track ends together with the reel on a bar boundary and sounds like a planned ending instead of breaking off mid-phrase:

• a long track loses whole bars from its middle and finishes on its real ending;
• a short track repeats its tail from a bar boundary up to the end of the reel;
• the join is smooth, and the track start you picked is kept.

Free and with no AI — it works from the track's own beats. The preview and the finished reel sound the same, on the web and in the app. Remove fit brings the whole track back.

Dub into another language#

Dub into another language is a new item in the Translate sheet. Pick a language and one of four voices — Slicer translates the reel's speech and voices it with an AI voice. You get a copy of the project with the same sources:

• the voice-over is on the Voice track, in time with the original; long phrases play a little faster or move into pauses;
• captions, the hook, plaques and titles are in the new language;
• the original speech is muted, and the rest of the clips' sound plays quieter in the background.

The price is at cost: the translation plus the seconds of voice-over, shown before you start. It usually takes 1–3 minutes, and you can keep editing. If dubbing fails, the tokens are refunded. The author's voice is not cloned. The speech must be transcribed (Captions from speech).

Fit to length · Dub into another language · Operation prices

Picture in picture#

The “+” on the strip has a new item, Picture in picture: a photo or video sits in a small window over the reel. An existing insert becomes a window with As window and goes back with Full frame.

• Edit the window right on the preview: drag it, the corner handle resizes, the circle on top rotates; pinch on iPhone and iPad, the mouse wheel on the web.
• The Window panel: Rectangle or Circle, Size, Corners, Rotation, Opacity, Border with a color, and Sound for videos.
• Keyframe — and the window glides, grows and rotates between keyframes.
• Up to 10 windows per reel. Free and without AI.

Remove object#

Video fragments up to 10 seconds long have a Remove object button (beta): draw a box around a stray item or a passer-by on a frame, and AI removes it across the whole fragment and fills in the background. Bring the object back restores the original video. The price is at cost, ≈0.93 tokens per second of the fragment, shown on the button before you start. On a busy background the result may not come out perfect.

AI access to files#

Some AI models need a file by link. Before such an operation Slicer asks for permission: Allow once, Allow for this project or Don’t allow. A link opens only the piece that's needed (for object removal — a piece of the fragment and the object mask), is valid for an hour at most and is revoked right after processing. The project's Settings have an AI access to files block: Sources, Own media and Finished video toggles and an Issued links log with a Revoke all links button. Turning access off revokes active links at once.

AI upscaling to 4K#

At 4K resolution the export window now has an Upscaling row: Standard stretches the frame for free; “AI” draws in fine detail at cost — ≈1.2 tokens per second of the reel. Such a reel is rendered by the server — in the app the Where to render choice is inactive then. If the AI can't manage, the reel is upscaled the standard way and the reserve is returned.

API 1.6.0 and MCP#

A new edit operation, set_pip: an insert as a window over the reel or back to full frame — position, size, rotation, opacity, shape, corners, border and sound. It works in POST /v1/projects/{id}/edits, the MCP tool edit_project, the AI assistant and the director in the chat.

Picture in picture · Remove object · AI access to files · AI upscaling to 4K · Editing in the API

Help where you work#

Every screen header has a “?”, and sheets have How does this work? at the bottom. Help opens right at the relevant article and section: from the editor's “+” sheet — about tracks, from the Speed panel — about clip speed. Editor tool panels have a “?” next to ✓, and setting hints have Learn more in Help.

• On a phone help takes the whole screen; on a computer, iPad and Mac it's a panel on the right taking almost half of the window, with no dimming: read and do at the same time.
• At the top — Search help across all the documentation as you type; a result opens the article at the found section.
• The Help home — Start here, Frequently asked questions, Topics and For developers.

Details — Help in the app.

In the docs#

• A new Voice recording section in the audio guide.

Color#

• LUT — a new clip tool: ten built-in looks — from Teal & orange to B&W film — with adjustable strength. The LUT goes on top of the filter and the manual settings.
• My LUTs — your own .cube files from DaVinci Resolve, Premiere or Final Cut. Save as LUT turns a clip's whole color into one .cube you can use in any project, download or share.
• Auto in Adjust balances a clip's color from its frames, and Match all to this matches the reel's other clips to it. No AI, and free.
• Adjust has two more sliders: Exposure and Tint.

Library#

Every editor preset in one panel: color and LUTs, effects, transitions, animations, six title styles, elements and sounds. A click puts the preset on the clip at the playhead or at the playhead itself; on the web, iPad and Mac you can drag a tile right onto the clip you want on the timeline.

Sounds#

Eight sounds — Whoosh, Pop, Ding, Click, Swipe, Riser, Boom and Chime — you can place yourself at any moment of the reel, replace, move and delete.

Fixes#

• In the app, the paused preview now shows a new filter, Adjust and LUT right away — without starting playback.

Public API 1.4.0#

POST /v1/projects/{id}/auto-color and the MCP tool auto_color; in operations — adjust and lut for tune_block and set_all_color, and the sounds add_sound, update_sound, remove_sound.

Color and LUTs · Library · Sounds · Auto color via the API

Where AI runs#

The Captions from speech and Remove background sheets have a new Where AI runs row:

• On device · free — your phone, tablet or Mac recognizes speech and separates the person from the background with Apple's models;
• In the cloud · price — an AI service at cost, as before.

The result is the same: word-by-word captions on the clips, the person on a blurred frame or a colored background. Progress shows above the preview — Recognizing speech on the device or Separating the person from the background with a percentage. If it fails on the device, the sheet opens again — try again or pick the cloud.

The preselected option comes from the new Profile → Where AI runs card. The device handles videos uploaded from it, and the language has to be set — Detect is cloud-only. On the web AI runs in the cloud, and the device option is labeled In the app.

In the project history such work is marked on the device.

Captions from speech · Remove background · Where AI runs

Where to render#

The export window in the app has a new Rendering section with the Where to render row:

• On this device — your phone, tablet or Mac renders the reel right away, with no queue. On iOS 26 and later you can minimize the app and the export keeps going in the background.
• On the server — as before: Slicer renders the reel, and you can close the app.

A reel rendered on the device looks the same as one rendered by the server: the same captions, color, transitions and effects. If rendering on the device fails, the export carries on — the server renders the reel with the same settings and publications.

The export card shows what is happening: Preparing the job, Downloading materials, Rendering on the device, Uploading the reel. In the browser the server still renders the reel, and the “?” next to the row explains rendering in the app.

Fixes#

• A new export shows Queued right away instead of the previous run's stage.

Where to render · Retry and stop

Talk about your reel#

Chat with the director opens from the project header. Ask why the reel was put together this way, or ask for a change: make it more dynamic, drop a part, change the order. The director edits the current montage with the same tools you use — the edits show up in the editor at once and can be undone. If a remount with new settings is needed, the director offers it and starts it only after you confirm.

Voice, files and the answer model#

• Voice messages — up to 5 minutes.
• Attachments “+” — photos, videos and audio: the file becomes a source of the project right away, and the director can put it into the reel.
• The answer model is picked under the input field; your choice is remembered.

What it costs#

A director's answer costs what the model's work costs. Under the field you can see how many tokens the conversation used and what it cost. Remounting the reel is separate and only with your consent.

Chat with the director · How the director works

Timeline#

• Marker — a coloured note on the ruler: it moves with its clip, and clip edges and elements snap to markers. Keys M, Shift+M and Alt+M — on the web and on an iPad or Mac with a keyboard.
• Copy look and Paste look — speed, sound, colour, effect, transition, animations, mask, background and framing from one clip to another or to all of them at once (⌥⌘C and ⌥⌘V).
• Slip — pick a different part of the same source for a B-roll clip without changing its length or place in the reel.

Picture and sound#

• Stabilize is stronger: two passes and a slight zoom instead of mirrored edges.
• A cover with a title — styled like the hook or as a callout; Cover as the first frame — for networks where you can't pick a cover.
• Fade in and Fade out for a clip's sound and for music.
• Colour by speaker — captions of two speakers in different colours.

12 new transitions#

Slides and wipes in all four directions, Circle open and circle close, zoom in, Pixelate, Radial wipe and Noise dissolve. The preview draws them the same way as the finished reel, on the web and in the app.

AI director#

The director edits with more restraint: clean cuts at natural speed, speech never sped up, no freeze frames, rare accents with an explanation. The three versions of a reel are ready sooner.

Fixes#

• In the app the preview no longer starts by itself when you open the editor, cuts don't jump or show bits of the source during playback, and the timeline scrolls smoothly.
• “None” for the music fade-in and fade-out now means no fade in the finished reel too.

Public API 1.3.0#

Markers, clip fades, apply_look, slip_block, set_cover and 12 new transition kinds in the operations catalog.

Markers · Clips · Transitions · Format and cover · Editing via the API

Edit it myself#

“New project” now asks how you want to edit: With the AI director, as before, or Edit it myself. With the second option your materials go onto the timeline in order, the project opens right away and stays a draft until the first export. Editing by hand, exporting and publishing are free and have no watermark. You can still call the director later: Edit with the AI director in the project's Settings.

Captions from speech and cutting pauses#

In a project without the director, captions appear with Captions from speech: Slicer recognizes the speech in the sources you pick, and the reel's elements stay where they are. Cut pauses removes silence longer than the threshold you choose and shows right away how much shorter the reel gets.

Titles and elements at the playhead#

A Title is free text over the video: font, size, colour, background, position and animations, including a typewriter effect. The hook, callouts, emoji and zooms can be added by hand at the playhead. Elements are tied to the reel's time and survive reordering, trimming and deleting clips.

AI at cost#

The AI assistant, translation, AI effect, background removal, avatar, voice isolation and speech recognition cost what the model's work costs. You see the price before starting and are charged no more than that. A montage by the AI director still costs 24 tokens.

Public API 1.1–1.2 and MCP#

Projects without the director (mode: "manual", draft status), elements placed by reel time, one catalog of edit operations (including titles and add_source), AI operation prices at cost. In MCP — create_project with mode and the edit_myself prompt.

Projects without AI · Captions from speech · Titles · Pricing · Editing via the API

Getting started with Slicer is now easier. New users go through a short path:

1. Intro — four slides about what Slicer does: editing raw footage, three cuts from the director, captions and camera, the editor and publishing.
2. Sign-up — with email and password or with VK ID, Google and Apple.
3. Free start or a pack — the first project is free, and a pack can be bought right away or later in Balance.
4. First project — what to prepare and a button that opens the New project wizard.

The onboarding step is stored in the account: if you start on your phone and continue in the browser, Slicer will not show what you have already done. Users who signed up earlier do not see the onboarding.

We also launched slicer.media with an overview of features and prices — from there you can go straight to sign-up.

Step 1. Sign-up and intro

Everything you do in Slicer by hand can now be done programmatically or handed to an AI agent.

Public API 1.0#

A REST API at https://api.slicer.media/v1:

• Uploads — in parts, as a whole in one request, or from a public link.
• Montage and price — the price is always named before the start and confirmed explicitly; without confirmation no tokens are spent.
• Projects — status, stages, versions, picking a version, history, downloading the reel and its cover.
• Edits — operations on the reel's text, clips and elements, or replacing the document; the AI assistant and caption translation.
• Export and publishing — the file profile and publishing to connected social accounts.
• Webhooks — signed notifications about finished uploads, reels, exports and publications.
• An OpenAPI 3.1 specification, errors with clear codes, per-token rate limits, idempotency, cursor pagination.

App-only for now: buying packs, connecting social accounts, background removal, voice isolation, AI avatar, your own media and music in the editor — they will come in later API versions.

MCP server#

The server at https://api.slicer.media/mcp connects to Claude, Claude Code, Cursor, VS Code, Codex, ChatGPT and other clients: ask the agent to make a reel from files, edit it or publish it. Sign in through the browser (OAuth) or with an API token. Paid tools require confirming the price, and each connection has a spending limit.

“For developers” in your profile#

A new screen in the profile: API tokens with scopes, expiry and spending limits, connected apps, the MCP server address and webhooks.

API overview · MCP server · For developers

Three cuts for one price#

The director makes the main cut, reviews the finished result and makes two more — with a different script, pacing and effects. When the montage is ready, the Pick a version sheet opens: watch them and pick the best. The others stay in the project history. All of it for the price of one montage, 24 tokens.

Smart cut — a group of settings#

The Smart cut row opens the Direction sheet:

• Video type: auto, talking head, story / vlog, product / review, aesthetic + music.
• Pace: calm, balanced, dynamic.
• Effects and transitions the director adds itself: none, light, bold.
• AI voiceover — for product footage and videos without speech: the AI writes a script from your shots and voices it with one of four voices.

Notes for the director#

On the Media step and in the project's Settings — a field of up to 600 characters: describe the mood, pace, what to show first and how to end. There are ready-made suggestions: Fast, no pauses, Focus on price and benefits, Calm, to music, With humor. Notes are stored encrypted.

The run card#

A stage that was not needed in a run is marked not needed and costs nothing.

Pick one of three cuts · Smart cut and notes

The web matches the app#

Everything that appeared in the app is now in the browser too: Projects, the new project wizard, the three-section project, history and versions, picking one of three cuts, the export window, social accounts and publications. The web editor is the same as in the app, with every tool, and is comfortable to drive with a mouse and keyboard:

• the wheel scrolls the timeline, Ctrl or ⌘ + wheel zooms;
• drag clips and elements, double-click a caption phrase to edit its text;
• keyboard shortcuts: Space, ← →, ⌘ZCtrlZ, ⌘BCtrlB and more;
• drag files straight into the editor and the wizard.

Old reel addresses lead to projects.

Wide screen#

From 960 px — on a computer, an iPad in landscape and a window on a Mac — a sidebar replaces the header and the bottom tabs, projects show as a grid of covers, sheets open as a dialog in the middle, and export and the wizard's first step use two columns. In the editor, the AI, Elements and Media panels open next to the preview.

iPad and Mac#

The Slicer app runs on iPad in all orientations, and on a Mac — as the iPad version from the App Store in a resizable window. With a keyboard, the editor supports the same shortcuts as the web, and the timeline scrolls with the trackpad.

The app for iPhone, iPad and Mac · Gestures and keyboard shortcuts

Sign-in and account#

• Email and password with a confirmation code by email, password reset.
• Sign-in with VK ID, Google and Apple. Accounts are linked by a verified email.
• Sign-in and security in the profile: change email and password, sign-in methods, a list of devices, Sign out of all other devices.
• Account deletion — right in the app and on the web.

Payments#

• Token packs are paid via YooKassa: on the web — in the same tab, in the app — in the system browser with a return to Balance.
• Tokens are credited only after the payment system confirms the payment; Slicer does not see card details.

Security#

• Files — sources, reels, covers — are stored encrypted, each file with its own key.
• Sensitive data — reel text, titles, notes, personal data — is encrypted too.
• The Slicer team does not see your files in plain form; every exceptional access is recorded in a log that cannot be altered unnoticed.

Sign-in and account · Balance and packs · Security

The app editor became a full editing desk: a full-width timeline, tools for the selection at the bottom, panels over the timeline. Every tool previously marked “soon” is done. The editor reached the web on 16.10.2026.

Clip#

• Speed from 0.25× to 4× without changing the voice pitch, and volume.
• Effects: shake, pulse, flash, cinema, vignette, film grain, blur, glitch, VHS.
• Transitions: fade, flash, zoom, dissolve, slide, wipe, blur.
• Color: filters and Adjust — brightness, contrast, saturation, temperature.
• Frame: in and out animations, mask, background, freeze frame, replacing a clip, reverse, stabilization, keyframes, frame-precise trimming.

Audio and music#

• Reduce noise, enhance voice, isolate the voice from music and noise, extract audio from a video.
• Music beats and snapping cuts between clips without speech to the beat.

AI#

• Caption translation into 11 languages with timing preserved — Translate costs ≈3 tokens, only if the translation works.
• Remove background — a person on a blurred background or a color; the price is shown before you start.
• AI avatar — the person in a photo says your text and is placed in the reel as a clip with captions.

The whole reel#

• Aspect ratio: 9:16, 4:5, 1:1 and 16:9 — framing and graphics rearrange themselves.
• Cover — any frame of the reel.

Editor overview

Export#

A full-screen export window: the Social, Master file, 4K for YouTube and Small file presets and manual settings — MP4 or MOV, H.264 or HEVC, 720p to 4K, 24/25/30 fps, quality, 8- or 10-bit, audio and loudness. The summary shows the frame size, bitrate and estimated file size right away.

Auto-publishing#

• Connect Instagram, YouTube, TikTok and VK Video — and the reel is published right after export.
• The AI drafts the title and the caption with hashtags; a Made with AI label for reels with an AI avatar and effects.
• Settings for each platform, including all of TikTok's rules for apps.
• Publication statuses in the project header: uploading, processing, Published with a link; retry and cancel.

The run card#

• Progress by the stages of the current run: an export goes from 0 to 100% rather than starting halfway.
• The time of each stage, how much has passed and how much is left.
• What is happening with the tokens: reserved, charged, returned or free; each stage's share of the price.
• Stopping with confirmation. A failed export keeps the previous reel, and retrying a free run is free.

Export · Publishing to social media

A big app update: Slicer is now organized around projects, and the montage accepts more than video. All of this reached the web on 16.10.2026 — see The web like the app.

Projects#

• Projects is the home screen. Search by title, filter by status, sorting, a long-press menu.
• A project in three sections: Media — all source files, Settings — title, style and montage settings, Editor — editing the reel. A style changed in Settings shows up in the Editor right away.
• Project history. A log of all work on the reel and its versions — the last 10 can be watched and made current.

The “New project” wizard#

Three steps with a stepper: style → captions and options → media and launch. Settings are full-width rows, each with a “?” hint. Uploads and the current step are kept even if you close the wizard.

Photos, music and voice-overs#

• Photos appear where they are mentioned, or as their own shot with a slow push-in.
• Music plays underneath and ducks under the voice; in a reel without speech it takes the lead.
• A voice-over becomes the narration, and the director picks visuals for it from your videos and photos.
• Slicer decides by itself which audio is music and which is a voice-over.

A more precise montage#

• The director looks at more frames and gets a single script: what is said and what is seen.
• Meaning over coverage: retakes are dropped, and the reel ends on a finished sentence.
• Smooth camera motion — push-in, pull-out and punch to the face, the same in the preview and the finished reel.
• 3D emoji in a new style.

Project list · Tutorial: your first video

B-roll without speech#

Previously the reel was built mostly from speech. Now the director understands what happens in shots without words and places such B-roll in the reel by meaning. Every uploaded video is used — nothing is lost, even if it has no speech.

AI effects from a prompt#

• Describe an effect in words — “make it snow” — and a neural network changes the video in the chosen range.
• The range is 2 to 30 seconds, or the whole reel if it is shorter.
• The price is ≈2.2 tokens per second of the range, shown on the button before you start.

AI effect from a prompt

Slicer is out as an iPhone app. It is the same service as in the browser — the same account, projects, montage and editor — but with gestures and the phone's features.

What's new#

• Media from your library. Pick videos and photos straight from Photos or from Files.
• Gestures in the editor. Reorder clips and move elements with a long press, auto-scroll at the screen edge, pinch to zoom the timeline.
• Precise trimming. A clip edge can be nudged frame by frame and snaps to words.
• Moving elements by words. Cards, callouts, push-ins, inserts and emoji move to other words.
• Share and the library. The finished reel is saved to Photos and sent through the system menu.
• Haptics on key actions.

The app for iPhone, iPad and Mac

A finished reel can now be polished before export. The editor preview is assembled the same way as the final file, so you see exactly what you get.

What's new#

• Captions and text. Fix words in captions, bring back cut parts, change highlights.
• A multi-track timeline. Video, caption and element tracks, trimming edges that snap to words, splitting and reordering clips.
• Framing and elements on the preview. Set a clip's framing and move captions and elements right on the preview.
• AI assistant. Describe an edit in words — “make it more dynamic and cut to 40 seconds” — and the assistant edits the reel. A command costs ≈2 tokens, and any edit can be undone.
• Your own media and music. Insert your own photos and videos over the frame and add music that ducks under the voice by itself.
• My templates. Save a style you like and apply it to new reels.

Export#

Export from the editor is free. Re-exporting is faster: only the changed clips are rendered again.

A new look#

Slicer got its logo — a 9:16 frame with a wedge cut — and the “marker” accent color.

Editor overview · AI assistant

The first Slicer release. Upload one or more videos from your phone, pick a style and tap Create reel — in a few minutes you have a vertical reel for Reels, Shorts and TikTok.

What it does#

• Cutting. The AI removes pauses, slips, retakes and filler words and keeps the strongest moments.
• 9:16 framing. Horizontal videos are reframed to vertical, the speaker stays in the shot; key phrases get push-ins and framing changes.
• Word-by-word captions with highlights in the chosen font and color. Style templates, fonts and accent colors.
• Graphics and sound. Emoji, text callouts, media cards with illustrations, subtle sound effects on cuts, loudness normalization.
• A hook. If you like, the AI writes a catchy title for the first seconds.

Payments#

• Work is paid with tokens; a reel costs 24 tokens.
• Your first reel is free, with a watermark. After your first payment it is re-rendered without the watermark for free.
• If a montage fails, the tokens return to your balance.

Reliability#

If an AI service is unavailable, the montage does not fail — it runs in a simpler way, for example cutting by pauses or framing at the center, and honestly marks the stage.

What is Slicer · Quickstart