The most popular open-source video editor for the Linux desktop is back with another update . Shotcut 25.10 is now available and comes packed with features aimed directly at productivity: locally running artificial intelligence, HTML generators for stylish titles, and native screen recording that eliminates the need for external utilities.
After a beta testing period, the release as a stable version confirms that the project is progressing steadily. The standout new feature is text-to-speech (TTS) for notes and subtitles, along with a typewriter filter, an image/video generator from HTML, and an update to FFmpeg 8, not to mention a substantial package of bug fixes and compatibility improvements.
Overview: What's changing in Shotcut 25.10
The team has been incorporating AI features recently: first came transcription with Whisper (speech-to-text), and now the circle is completed with integrated, local text-to-speech . This combination opens up powerful workflows for narrating, subtitling, and making content accessible without leaving the editor.
Alongside AI, this release introduces creative tools supported by HTML and features requested by the community, such as cross-platform screen recording. The release also enhances the technical foundation with updated FFmpeg 8 and key libraries.
Shotcut 25.10 introduces local text-to-speech for notes and subtitles
Shotcut's TTS implementation relies on KokoroDoki and the Kokoro 82M model , an open-source speech synthesis engine that can operate in real time. It's worth noting that version 25.07 incorporated a speech-to-text model . Most importantly, everything happens on your computer: there's no data sent to the cloud, thus guaranteeing privacy and project control.
To speed up processing, KokoroDoki can use a CPU or NVIDIA GPU with CUDA . On machines with compatible graphics, you'll notice significantly faster generation times, which is especially welcome when iterating over many lines of subtitles or preparing long voiceovers.
Regarding voices, the current package is more comprehensive in English. American English has over twenty voices , British English offers around eight, and coverage for other languages ​​is more limited for now. It's expected that the range will expand in future revisions, but initially there's already enough material for mock-ups, prototypes, and final projects in English.
There's one technical detail to keep in mind: the integration is done using Docker . This adds an extra configuration step compared to other features, especially if you've never worked with containers before. It's not an insurmountable obstacle, but it's worth setting aside a few minutes to fine-tune everything.
Shotcut 25.10, generating from HTML and typewriter effect
Creativity expands with two complementary additions: a new "Image/Video from HTML" generator and a text effect that simulates a typewriter. The focus is clear: to bring browser capabilities to the timeline to produce titles, captions, and animations with transparent backgrounds and without any unusual plugins.
The HTML generator creates images or clips with transparency by default . To maintain stability, the generated videos are limited to 15 frames per second, a sufficient value for graphics and overlays that prevents bottlenecks. As soon as it finishes, the result automatically opens in the Origin viewer along with its HTML, so you can review or edit the code on the fly.
The "Text: Typewriter" effect comes with ready-to-use presets and others designed for customization. Among the included profiles are 3D image, elastic stroke, folded, gold metal , and party time , covering everything from solid image finishes to more playful animations for video. If you're comfortable with CSS and JS, you'll have the opportunity to create some very sophisticated effects.
- Multi-device Compatibility Code compiled from CodePen, ideal for reusing snippets and designs you already have in your stash.
- Render with a limit of 15 FPS for stability in browser-based generation.
- Creative possibilities with transparent backgrounds without complicated compositions.
Native and cross-platform screen recording
Another long-standing community request has finally arrived: integrated screen recording . Implementation varies depending on the environment, but in all cases, the most reliable native backend has been chosen to minimize friction.
- X11- Using FFmpeg as main backend.
- Wayland (GNOME): integration with the native capture/screencast tool.
- KDE Plasma: support in Spectacle for capture.
- AlternativeIf none of the above are present, then resort to OBS Studio as a wild card.
This eliminates the need to install additional utilities in most cases and streamlines the workflow : you record, edit, and export from the same application. For creators of tutorials, software reviews, or training sessions, this is an improvement that's noticeable from day one.
Shotcut 25.10 updates the multimedia engine: FFmpeg 8 and libraries
There's also movement under the hood. The application is moving to FFmpeg 8 , which means improvements in encoding/decoding, new filters, and broader compatibility with modern formats. On a practical level, this translates to more consistent exports and smoother playback of complex projects.
Essential components such as SVT-AV1, libaom, dav1d, libvpx, libwebp, and whisper.cpp are also being updated . These components affect AV1, VP9, ​​and WebP codecs, as well as the AI ​​transcription itself, completing the loop for more robust workflows.
Shotcut 25.10 interface and workflow improvements
Small details that add up: the Properties panel now includes the "Open with" and "Reload" options . These are modest but practical changes when you need to refresh externally edited media or quickly launch it with a system app.
The new "Image/Video from HTML" generator works alongside the "Text: Typewriter" filter, allowing you to combine HTML generation outside the timeline (as a media asset) with typewriter animation applied to an already inserted clip. Complete flexibility depending on the use case.
Shotcut 25.10 bug fixes and stability improvements
This release fixes several annoying problems that had been reported, similar to update 25.08 . For example, it corrects export failures when the name or path contained the ampersand (&) character , a more common issue than it might seem in shared projects.
This update also resolves an issue when launching the AppImage on Linux with AppImageLauncher installed, and fixes advanced keyframes in "Text: Rich" that weren't behaving correctly. Finally, it corrects a bug in the alpha channel when decoding Ut Video , which is crucial for those working with transparency in professional pipelines.
Installation and compatibility
To make life easier on Linux, the project distributes a universal AppImage that you can download and run without installing any dependencies. In practice, you just need to grant it execute permissions and you're good to go.
On the compatibility front, support for glibc 2.35 is confirmed in Ubuntu 22.04 LTS , so that LTS version is covered. Any modern distribution with a kernel higher than 5.10 should work without any problems, while on older systems you might need to adjust some libraries.
Use cases: from creators to educators and developers
For those who create content for social media or YouTube, this package is perfect. You can record your screen without leaving the editor , apply stylish text effects, and generate synthesized voices for intros, explanations, or calls to action without having to manually record every change.
- Video tutorials: integrated recording + TTS annotations for dynamic rhythms.
- PresentationsProfessional animated titles without paid plugins.
- AccessibilityBy combining Whisper (voice to text) and TTS you can improve subtitling and narration.
In education and training, the value is clear: recorded classes, video tutorials , and support materials with locally generated titles and voiceovers. Multiple English voice profiles allow for a more varied tone depending on the audience.
- Software demosNative application capture without relying on external tools.
- Visual documentationVideos with technical annotations and readable bursts of text.
- Interface testing: record and review workflows for quick feedback.
Quick comparison with the previous version
Compared to the previous version, the improvements are substantial. Native recording eliminates the need for additional solutions on most desktops, and the HTML+Typewriter combination greatly expands the options for titles and lower thirds.
- ✅ Integrated recording without going through OBS (except in alternative scenarios).
- ✅ Advanced text effects with HTML/CSS support and quality presets.
- ✅ Text to speech for notes and subtitles within Shotcut itself.
- ✅ ffmpeg 8 for better performance and current formats.
- ✅ Enhanced stability with fixes in common edge cases.
Requirements and recommended configuration
For everything to run smoothly, it's best to aim for a well-balanced system. 8 GB of RAM is sufficient , but if you're working on medium to large projects, you'll appreciate having 16 GB.
- CPU: at least 4 cores for effects and loose encodings.
- RAM8 GB minimum, 16 GB recommended.
- GPU: Compatible with OpenGL 3.3+; if you use CUDA, it will accelerate TTS with KokoroDoki.
- StorageSSD for media and cache, especially in projects with heavy codecs.
In project settings, it's a good idea to match the FPS and resolution to your final output to avoid unnecessary upscaling. Keeping proxies and preview optimized will prevent stuttering when adding filters and multiple tracks.
Workflows with HTML, CodePen, and Typewriter
If you're going to fully utilize the HTML generator, work on the design in your editor or on CodePen , compile/package the resources, and then paste them into the Shotcut template. This ensures that fonts, styles, and scripts are handled without external calls.
For text animations, alternate between the "Text: Typewriter" filter and external generation: when you want speed and something lightweight, the filter will do the trick; if you need complex finishes, the HTML flow will give you more control (and native transparency).
Notes on performance, privacy, and limits
The local approach to TTS has two major advantages: it protects your content and avoids latency or online service fees. In return, it requires setting up Docker and, for maximum speed, an NVIDIA GPU with CUDA capabilities.
The 15 FPS limit for HTML rendering isn't arbitrary: it stabilizes rendering when the browser engine generates video or images with transparencies and layers. In practice, it's more than enough for titles, transitions, and overlays.
Small details that make a difference
The "Open with" option in Properties makes it quick to edit a clip in an external tool and return to it without losing track, while "Reload" allows you to force a media update if the file has changed on disk. These two keys, plus the Source viewer that automatically opens HTML-generated content, make everyday tasks much smoother.
The combination of FFmpeg 8 and the updated libraries also results in a smoother experience: fewer surprises when importing/exporting and better handling of recent formats like AV1, along with finer decoding.
Those coming from previous versions will find an editor that has matured without losing its essence : free, cross-platform, and with technical decisions that favor working locally. The feeling is that the project is hitting the gas right where it needs to: performance, practical AI, and creative tools with gentle learning curves.
