Changelog

The text chat service now supports GPT-6 Luna, DeepSeek V4.1 Flash and Gemini 3.5 Flash, with Luna as the default. These choices support the upcoming desktop Chat workspace. Gemini Flash uses more credits than Luna and DeepSeek.

Conversation context is preserved across follow-ups, with bounded message sizes and reply lengths. The service rejects unsupported model choices and keeps provider credentials on the server. Private Admin usage reporting distinguishes accepted chat requests from completed desktop replies, copies and saves.

Chat now supports explicit Thinking and Non-thinking requests with GPT-6 Luna and DeepSeek V4.1 Flash. Non-thinking is the default for these models; Thinking uses low effort. Gemini 3.5 Flash requires Thinking.

Responses remain limited to the same budget model list and 2,048 output tokens. Studio credit billing includes reasoning tokens when enabled. Mode selection itself does not consume credits or add feature usage.

Chat now accepts images and document text with Luna, DeepSeek V4.1 Flash and Gemini 3.5 Flash. Files can be included in follow-up conversations while retaining the existing Thinking and Non-thinking controls.

The upcoming desktop preview reads PDF, Word and text documents locally and prepares images before sending. Files are included only when you send a message. Context limits and the same budget model list remain enforced; images and document text use Studio credits when sent. No premium-model fallback or additional file-parser service is used.

Chat now supports optional writing preferences with the same budget models and Thinking controls. Instructions are included only when enabled and apply to the next message. Replies keep the existing Studio credit billing and response limits.

The upcoming local desktop preview adds Natural writing in Settings, with editable instructions, Save instructions and Reset to default. The default combines a conversational tone and practical examples with full forms such as “I am,” short paragraphs, numbered steps and no dash punctuation. The preference starts off and saves on your device. Public desktop downloads are unchanged.

The homepage and footer now include our official Product Hunt badge. Visit the SmoothyEdit listing to follow the released desktop app’s launch and share feedback.

The Product Hunt badge now appears only in the footer, keeping the homepage focused on the editing tools and download buttons.

AI Photos now offers 15 image models. New choices include FLUX.3 Image, FLUX.2 Pro, Flex, Max and Klein, Seedream 5.0 Flash and 4.5, Grok Imagine Image 2.0 and Riverflow V2 Fast.

The available canvas shapes, resolutions and reference limits follow the selected model. Generation and edits use Studio credits, with image-based API charges included in the final cost. Selecting a model or attaching a picture does not spend credits.

The private Admin usage table includes feature totals and verified Free, Paid and Anonymous request counts. Each accepted image request counts once, separately from credit billing. Newly recorded activity starts with this update; no earlier Studio usage is invented. Desktop counters appear as supported app builds report them.

Removed the installation and revenue tab below the navigation across the site. These totals are no longer shared through a public stats endpoint; analytics remain available privately in Admin.

Admin → Usage now groups anonymous desktop installations by their latest reported app version, with installation totals, percentage share and activity over the past 30 days.

The existing app telemetry already reports versions, so this works with released builds without requiring an update. Each installation is counted once and moves to its new version on its next report. Offline installations retain their last reported version; installations with telemetry disabled are unavailable. These private counts represent installations rather than unique people.

• Studio’s tool buttons now use four compact columns on wider screens, with a distinct color for each tool and comfortable touch targets on smaller screens.
• Shorts markers identifies candidate ranges and sends markers to a connected Premiere timeline. You make the cuts, reframe and export the finished video.
• Hook ideas suggests opening moments and wording. You choose the opening and edit it into your video. The results explain this too.
• Condense plan, thumbnail, B-roll and sound-effect controls describe the suggestions they provide.
• Removed Add Music from Studio and its customization menu. Existing saved preferences skip that retired tool. The free desktop Audio Library remains available for choosing music.
• Account credit checks stop further cloud AI calls after the balance is exhausted. Repeated payment events no longer refill spent credits; previously issued balances keep their normal reset date.

• Studio Pro is now $5 USD per month, with 2,000 AI credits every 30 days.
• Free accounts still receive 100 AI credits every 30 days.
• Previously issued credits remain available until their normal reset; the new allowance applies to new balances and subsequent resets.
• Shorts provides timeline markers and hooks provides opening ideas. You choose and make the edit.

Admin → Usage now shows new installations per day, with 7-, 14-, and 30-day views, platform totals, a seven-day average and a daily numbers table.

Dates use Cairo time. Today is marked as incomplete and compared with yesterday after the same elapsed time; averages and weekly comparisons use completed days. Earlier dates without recorded history are not filled with invented activity.

The chart uses existing anonymous first app reports. It counts each installation once, independently of download requests, launches and app updates. No new desktop release is required, and the data remains private.

The homepage and SmoothyApp page now preview the two libraries coming in 1.5.1:

• Stock Footage: search Pixabay, Commons, Internet Archive and NASA, preview a shot and save an MP4 or send it to Premiere’s Project panel.
• Audio Library: browse an unofficial YouTube Audio Library archive, preview artist-made tracks, keep Favorites and start with Staff picks selected by the team.
• Updated mockups: explore both library tabs, filter sample shots and music, try Favorites and Staff picks, and see the import actions. These previews use sample media and illustrative picks.
• A wider workspace: the SmoothyApp introduction sits above a wide interactive mockup, with more room to explore the tools.

The homepage, SmoothyApp page and Multicam page include the updated owner notice, Discord/social links, source-clip and nested-sequence guidance, and the reminder to review both speakers and audio sync. Multicam’s 1.5.1 duplicate-timeline workflow is being tested, with the latest development fix preserving linked microphone audio and effects on camera clips.

Track licenses in the current music archive remain unverified. Check the track in YouTube Studio before publishing, including its attribution requirements.

Read the 1.5.1 feature preview for the workflow and details, or explore both libraries.

1.5.1 is still in development. Public desktop downloads remain 1.5.0.

Release update — October 4: Mac 1.5.1 is now available, signed and notarized. Windows remains on 1.5.0 while its update is being built. See the 1.5.1 release notes.

A compact strip was added below the navigation for aggregate installation and revenue totals, including on mobile.

This strip and its public stats endpoint were removed on October 5, 2026. Analytics remain available privately in Admin.

The Multicam warning on the homepage, SmoothyApp page and Multicam page is now a compact, softer notice. Open Details to read the full update, support links and timeline review guidance.

The close button still remembers dismissal across pages and reloads.

Removed the large gap between the Multicam update and the page introduction on SmoothyApp, Multicam and the homepage, on both desktop and mobile.

Moved the install and revenue totals into a small tab attached to the bottom-right of the navbar, on desktop and mobile.

SmoothyApp 1.5.1 is available for Apple Silicon Macs, signed with Developer ID and notarized by Apple. Download for Mac or read the GitHub release notes.

Fixed

• Multicam: source clips and nested sequences work on a duplicate of your original timeline. The chosen angle stays on the top selected camera track, with the reverse footage visible and selectable underneath for fine-tuning. Existing clip effects and continuous microphone audio are retained.
• Viewing another sequence during analysis no longer interrupts the captured job. Microphone exports use an app-created job folder, avoiding protected macOS temporary folders.
• The original timeline stays intact. Failed or incomplete camera edits are reported for review instead of being marked complete. Start from your original timeline and review both speakers and audio sync before using the result. Multicam remains beta.

New and improved

• Stock Footage: search Pixabay, Commons, Internet Archive and NASA, preview shots, save MP4s and import into Premiere’s Project panel.
• Audio Library: preview artist-made music, filter by title, artist, genre or mood, keep Favorites and Saved MP3s, and explore Staff picks. Local favorites work without an account; anonymous favorite sharing can be switched off.
• Licenses & credits: retain source notes and supplied attribution. The music archive is unofficial and does not supply verified track licenses; check the track in YouTube Studio before publishing.
• Community: LinkedIn and personal website links in Settings, plus an Instagram shortcut beside Discord.
• The website now includes the new library page and article, an interactive workspace demo, the compact dismissible Multicam notice, and installation/revenue totals beneath the navigation.

Windows: the current installer and update files remain 1.5.0 until the Windows 1.5.1 build is ready. The new libraries and Multicam fixes require 1.5.1.

The install and revenue totals now hang from the navbar’s bottom-left instead of sitting on top of it, with a softer lime tint so the tab reads as part of the navigation on desktop and mobile.

We’ve added a shared Pixabay search service for the next SmoothyApp release. Official app builds will include Pixabay in Stock Footage and All sources without asking you to create an account or supply an API key.

The service keeps credentials on the server, caches search results and limits requests. Clips retain their source links and usage terms. The current public desktop release remains 1.5.0.

Admins can search the audio archive, review aggregate favorites, and choose Staff picks. The service counts one active favorite per track and anonymous app installation, including free users, and removes its contribution when a track is unstarred.

The desktop interface is being tested in the local 1.5.1 build. Public desktop downloads remain 1.5.0; community favorites start arriving as users receive a compatible build. The privacy policy now describes anonymous favorite sharing and local-only favorites.

Multicam wasn’t working correctly in the previous update, 1.5.0. I’m sorry about that. A fix is being tested for 1.5.1, including source clips and nested sequences. Public desktop downloads are still 1.5.0.

The homepage, SmoothyApp page and Multicam page now show a dismissible update notice. Close it once and it stays hidden on future visits in the same browser.

If you’re having trouble, message me on Discord, X or Instagram.

Your next shot and soundtrack belong close to your edit. The new Stock Footage & Audio Library page previews two free tools coming in SmoothyApp 1.5.1.

• Stock Footage: search Pixabay, Wikimedia Commons, Internet Archive and NASA together, preview clips in the app, save an MP4 and send it to Premiere’s Project panel.
• Audio Library: browse an unofficial YouTube Audio Library archive, preview artist-made music, star your favorites and start with music curated by the team in Staff picks.
• One workspace: discover media alongside your editing tools, then place imported files on your timeline. Source links and license details help you check each item’s terms.

The current music archive does not supply verified track licenses. Check the track in YouTube Studio before publishing, including whether attribution is required.

1.5.1 is still in development. Public desktop downloads remain 1.5.0; this is a preview of the upcoming release.

Release update — October 4: Mac 1.5.1 is now available, signed and notarized. Windows remains on 1.5.0 while its update is being built. See the 1.5.1 release notes.

New

• Admin Usage has an Assets activity summary for successful PNG exports (including Save As), sends to Premiere and total actions. Each successful image counts once per action, and Assets appears as one tool in the ranking.
• Tracking is prepared in the desktop source for the next app update. Published 1.5 does not report Assets actions, so historical usage cannot be recovered. Events use the existing anonymous install ID and action names.

Fixed

• The Tool runs total now counts tool events instead of including launches and heartbeat pings.

Improved

• Usage first. Usage is now the first admin tab and opens automatically when you visit the admin panel.

Improved

• Added “Free forever. No account needed.” to the home and plugin page introductions, with “No sign-up. No credit card. No trial.” beside the downloads.
• Removed the version number from the SmoothyApp page heading and title. Release notes still identify the app version.
• Studio AI remains clearly optional.

Improved

• The SmoothyApp download announcement now clearly confirms 1.5 availability on both Mac and Windows, with a link to explore the new features.

SmoothyApp 1.5 is available for Apple Silicon Macs and Windows x64. The Mac build is signed and notarized. Download for Mac or Windows. Restart Premiere and reopen the SmoothyEdit panel after upgrading; sign in again to use Studio with the updated secure connection.

New

• SVG importer & Assets: convert SVGs and images into PNGs, preserve transparency, set a proportional output size, paste from the clipboard, export multiple files, or send the selected PNG to Premiere.
• Edit caption text and cue timing. Find and replace across the transcript, with case and whole-word options and one-step undo.
• A collapsible sidebar that remembers its state, with colored tool navigation and simpler account and connection controls.
• Shorts sources for Premiere sequences, transcripts, audio/video files and YouTube transcripts, with Multiple shorts and Best part modes.

Improved

• SVG imports default to 1000px wide with proportional height, including batch exports. You can still choose a custom width.
• Direct caption controls for language, model, engine, line length, line count and duration, plus local reformatting.
• Shorts history search, separate title and description copying, cancellation and a history-save retry.
• Camera mappings persist, and each tool explains what its action does above the buttons.
• Refreshed homepage and plugin page, interactive 1.5 mockups, a dedicated Assets page, and new 1.5 guides.
• The Assets preview and sharing images use the Smoothy cup logo as the sample SVG.
• The Assets page and SVG guide explain the Illustrator/converter detour and how to paste copied images with Ctrl+V or Cmd+V before sending them to Premiere. Removed the Assets item from desktop and mobile top navigation and the New badge from its footer link.
• Clearer search titles, social previews, article metadata and links between feature pages and guides.

Fixed

• Signed Studio sessions and scoped relay tokens replace caller-supplied user IDs. Premiere bridge connections are local and authenticated, and downloaded Whisper engine archives are verified.

• Mac automatic updates include the ZIP payload alongside the DMG installer.

• Centered the Studio Shorts sign-in panel, including its heading, feature list, button and credit note. Small windows keep the panel scrollable.

• Settings keeps its title and close button below the Mac window controls while the options scroll separately. The first row stays visible when scrolling back to the top, and reopening Settings starts at the top.

• Improved source and sequence timing, selected microphone handling and mixed-frame-rate processing.

• Clearer per-file asset errors and proportional sizing during export.

New multicam and default silence-removal sequences do not copy existing effects and titles. Shorts analysis provides ideas and markers for review rather than a finished vertical edit.

• Browse editing guides by task, with an on-page contents list and related articles.
• Explore all editing tools and the new Studio overview.
• Corrected older guides to match the current desktop app and web dashboard, including local processing, Premiere sequence limits and the manual Resolve subtitle workflow.
• Removed missing image references and repaired links to retired product pages. Older moved URLs now redirect to the relevant current guide or product page.
• Improved search metadata, article information, canonical URLs and sitemap dates. Added an article feed.
• Account and app screens stay out of search results; public product pages and guides remain crawlable.

Fixed

• Fixed Shorts analysis failing with “Reasoning is mandatory for this endpoint and cannot be disabled.” Studio now allows the AI endpoint’s default reasoning behavior while returning the final result.

New

• Language picker. Choose Auto-detect, or pin the exact language from every one of the 99 languages Whisper supports (English, Spanish, French, German, Japanese, Chinese, Arabic, Hindi, and so on). Auto-detect figures out the spoken language for you.
• Multilingual models. The model list now includes Tiny, Base, Small and Medium in both their English-only and multilingual variants, alongside Large v3 Turbo. Pick a multilingual model to transcribe anything other than English.

Improved

• Clear guard rail. If a language is paired with an English-only (.en) model, the app says so and tells you to switch to a multilingual model instead of quietly producing English text.
• Better progress text. While transcribing, the status line names the language (or says it is detecting one) and the engine in use.

New

• Multilingual captions. Choose a spoken language or let Smoothy detect it with a multilingual Whisper model. The available choices include many languages outside Premiere Pro’s built-in Speech to Text list.
• Signed Mac installer. The macOS 1.4 download is signed with Developer ID and notarized by Apple.

Improved

• The SmoothyApp preview now shows the 1.4 Captions settings, including model, language, and engine controls.
• The free DaVinci Resolve subtitle guide now has its own section with a simple SRT timeline preview.

The SmoothyApp page now walks through the free DaVinci Resolve subtitle workflow in three visual steps: generate captions in SmoothyEdit, save the SRT, and import it as a subtitle track in Resolve. The plugin availability note now appears at the bottom of the page.

Improved

• Two-column hero. The app mockup now sits on the left with the headline, description and download buttons on the right, so the first screen shows the product instead of a centered block of text.
• Clearer status notes. The “active development” and DaVinci Resolve notices were rewritten to be shorter and more direct.
• Proper download icons. The macOS button now carries the Apple mark alongside the existing Windows logo.

Improved

• Added icons to the desktop app features and a simple timeline preview for the Resolve subtitle workflow.
• Moved the active development note into the page content and simplified the DaVinci Resolve coming-soon message.
• Removed outdated Mac installation advice, the older compatibility notice, and Resolve email signup forms.

Improved

• Honest positioning everywhere. The homepage, page titles, footer and FAQs now say “free and open source app, optional Studio cloud AI” instead of implying the whole product is free.
• Rewritten pricing page. Two products instead of a confusing Free/Premium split: the App is $0 forever (multicam, silence removal, captions, compressor — MIT, no account), and Studio is the optional cloud add-on, free to start with a $10/month Pro plan.
• No more fixed “16 tools” claim. Studio is described as a cloud add-on that grows, so adding the next cloud tool won’t make the FAQ wrong.
• Best Shorts is clearly labelled. It’s marked “Studio” on the app page and in the desktop sidebar instead of being listed among the free local tools.
• Studio naming unified. The dashboard, account pages and desktop app now call the paid tier “Studio” rather than “Premium”/“Studio Pro”.

The desktop app keeps its local tools free and badged-free; paid cloud tools live under a single Studio group in the sidebar.

New

• Auto-multicam landing page. /features/auto-multicam now explains how speaker-based camera switching works, who it’s for, and why it’s free — with a side-by-side comparison against paid auto-multicam plugins and doing it by hand.
• New guide: The Best Free Auto-Multicam Plugin for Premiere Pro (2026). An honest look at the free, paid and cloud options, plus setup steps and best practices.

Improved

• The SmoothyApp and auto-multicam pages now answer questions in structured FAQ markup, so search engines and AI assistants can read the answers directly.
• Sitemap updated to include the app and auto-multicam pages.

A quick fix for anyone hitting the error below while generating captions.

Fixed

• Captions no longer fail with “ENOENT: no such file or directory … smoothyedit-whisper-…json”. The transcription output file was being deleted a moment before the app read it, so generation aborted at the very end. The app now reads the result before cleaning up, and captions are produced as expected.

A big quality pass on captions and updates.

New

• Transcribe any audio or video file. The Captions tab now has a source switch, so you can generate subtitles from a plain file on your computer — no Premiere sequence required. Handy if you just want an SRT to import into DaVinci Resolve.
• Cancel a transcription mid-run. A Cancel button now appears while captions are generating, so a long file no longer means waiting it out.
• A proper update panel. Updates now show up in Settings with the full “what’s new” notes and a Restart & Update button, so you can see what a version fixes or adds before installing.
• Join our Discord from the app’s sidebar and Settings.

Improved

• Update checks are more reliable. Fixed cases where the update button stayed disabled or hidden after a download, and where clicking it could fire twice.

Fixed

• Captions no longer stop partway through a long recording. A timestamp glitch could make the subtitle timings collapse after a few minutes, so players and Premiere appeared to cut the file off. Timings are now guaranteed to be valid and in order.
• “Send to Premiere” no longer throws os is not defined.

A note on DaVinci Resolve and UXP

• The DaVinci Resolve bridge and the Premiere UXP extension have been removed for now so we can focus on making the core app rock solid. They’ll return as soon as they’re ready.
• Using DaVinci Resolve Free? You can still make subtitles today — see the free guide.

Fixed

• The update bar’s “Restart to Update” button now works. It previously had no click handler at all — clicking it did nothing, which looked exactly like being stuck in an update loop: the app kept telling you an update was ready (and after every restart it re-offered the same update) because the install was never actually triggered. The button now installs the update and relaunches into the new version.
• Stuck-update detection. If a downloaded update never installed (the running version is still older than the downloaded one), Smoothy now tells you explicitly in the update bar and Settings, and skips re-downloading the same update so the loop can’t repeat.

New

• Transcription engine switch. Settings → Captions now has an engine picker: Auto (default, uses the GPU when available), GPU (CUDA), or CPU (all cores). The CPU engine uses every thread your machine has, so it’s as fast as CPU can get.

Fixed

• Automatic GPU → CPU fallback. If the CUDA engine crashes or can’t start at runtime (driver problems, unsupported cards, VRAM exhaustion), Smoothy now retries once on CPU instead of failing the transcription.
• “Send to Premiere” no longer imports partial captions. Importing now uses Premiere’s dedicated caption-track API instead of generic clip insertion, so the full set of cues lands on the sequence every time. If Premiere still refuses, you get a clear message with the cue count.
• Sequence audio is now trimmed correctly. Sequences with trimmed clips transcribed from the wrong part of each source file (the clip’s in/out points weren’t reported). The stitched audio now matches the timeline exactly.
• More resilient Windows updater. Updates install and relaunch with isForceRunAfter, so a completed update can no longer leave you restarting on the old version.

This release rebuilds the transcription engine so captions actually use your GPU.

New

• Real GPU acceleration. Transcription now runs the whisper.cpp engine with a proper CUDA build (the CUDA runtime is bundled), so NVIDIA cards are used instead of quietly falling back to CPU. Apple Silicon uses Metal. Caption Settings shows which backend you’re on — for example “Running on GPU (CUDA)”.
• Pick your transcription model. Choose Tiny, Base, Small, Medium, or Large v3 Turbo in Caption Settings. The model downloads on first use and your choice is remembered for next time.
• Caption cleanup tools. “Capitalize All” and “Remove Punctuation” apply to every generated caption and rebuild the SRT/VTT instantly.

Improved

• Clearer progress. The app now tells you when it’s downloading the engine and shows exactly where it’s running, for example “Transcribing on CUDA…”.

Fixed

• GPU builds that silently fell back to CPU. The previous native binding targeted the wrong CUDA architecture and omitted the CUDA runtime; the engine is now downloaded and run directly.

New

• SmoothyEdit for macOS. The desktop app is now built and published for Apple silicon (arm64), matching the Windows feature set. Download it from the site; existing installs auto-update.

Improved

• Transcription engine switch (Settings → Captions): Auto (GPU if available), GPU (CUDA), or CPU (all cores). If the GPU engine fails or is slow, switch to CPU — Auto retries on CPU automatically.
• whisper.cpp CLI engine with real GPU acceleration and a model picker, so captions run locally and faster.

Fixed

• Reliable caption import into Premiere via createCaptionTrack (CEP host script v7) instead of flaky clip insertion.
• The update bar’s “Restart to Update” button now installs the update and relaunches, and the app warns you when a downloaded update failed to install instead of looping.

Captions got a big reliability pass on Windows. Here’s what changed.

Fixed

• Captions cover the entire sequence. Transcription previously read only the first clip on each audio track, so a one-hour podcast could come back as a 14-minute transcript. Audio is now built from every clip on the tracks you select, honoring each clip’s in/out points and its position on the timeline.
• Send to Premiere no longer fails with a “header error.” The caption file is now written by the desktop app (the same bytes as Save SRT) and handed to Premiere for import, instead of being rebuilt inside Premiere’s scripting engine with a broken Windows path.
• Audio track selection works. The track checkboxes in the Captions panel now control which tracks are transcribed.

Improved

• GPU acceleration is automatic. Transcription picks the best backend for your machine — CUDA on NVIDIA, Vulkan on AMD/Intel, and Metal on Apple Silicon — and falls back to multithreaded CPU only if no GPU is available. Expect a large speedup on start-to-finish time.
• The app now reports the detected backend and the extracted audio length in its logs, so a short transcript is never silent again.