ClipShip updates are shipped through the desktop app's built-in updater. When an update is available, the app shows an update prompt with a link back to this page.
If an update does not install correctly, or if a bug is still reproducible after updating, email [email protected] with the app version and what you were trying to do.
v1.0.45v1.0.44v1.0.43v1.0.42v1.0.41v1.0.40v1.0.39v1.0.38v1.0.37v1.0.36v1.0.35v1.0.34v1.0.33v1.0.32v1.0.31v1.0.30v1.0.29v1.0.28v1.0.27v1.0.26v1.0.25v1.0.24v1.0.23v1.0.22v1.0.21v1.0.20v1.0.19v1.0.18v1.0.17v1.0.16v1.0.15v1.0.14v1.0.13v1.0.12v1.0.11v1.0.10v1.0.9v1.0.8v1.0.7v1.0.6v1.0.5v1.0.4v1.0.3v1.0.1v1.0.0
v1.0.45: Spoken language, clip titles in your language and clearer reports
October 22, 2026
This update remembers the spoken language you choose, writes clip titles and post captions in your video's language, and tells you clearly when a video's speech can't be transcribed reliably.
Changes
- The Spoken language you choose on the Style step is now remembered the next time you open ClipShip.
- Clip titles, hooks and post captions are now written in the language spoken in your video instead of English.
- If ClipShip can't transcribe a video's speech reliably, for example in a language it doesn't support well yet, it now stops and says so instead of making clips with mixed-up captions.
- Problem reports sent from Help & Support now include the detected language and no longer include your Windows user name.
- On Mac, problem reports now explain more when transcription stops unexpectedly.
Notes
- Your source video stays on your computer. A support report includes your app version, system details and recent activity, never your video, audio, transcript or API key.
- Pricing, subscriptions, referrals, device limits, and existing customer access are unchanged.
v1.0.44: Caption exports, older graphics drivers and clearer AI errors
October 18, 2026
This update fixes captioned exports failing on some PCs, keeps local AI working when the graphics card engine can't start, and explains AI account and speed problems clearly.
Changes
- Fixes captioned exports that failed every time on some PCs because ClipShip picked up an older copy of a library installed by another program.
- If the graphics card AI engine can't start (usually an old NVIDIA driver), local AI now switches to the processor instead of stopping.
- When your OpenAI, OpenRouter or Anthropic account is out of API credit, ClipShip now says so instead of reporting a rate limit, and real rate limits are retried after a short wait.
- If local AI is running on the processor and taking a long time, ClipShip now tells you why and how to speed it up.
- Problem reports sent from Help & Support now include the details we need to find the cause faster.
Notes
- Your source video stays on your computer. A support report includes your app version, system details and recent activity, never your video, audio, transcript or API key.
- Pricing, subscriptions, referrals, device limits, and existing customer access are unchanged.
v1.0.43: More clips from smaller AI models, damaged videos and support tickets
October 16, 2026
This update stops smaller local AI models from finishing with zero clips, keeps the readable part of damaged videos, and lets you send a problem report from inside the app.
Changes
- Smaller local AI models (Qwen 1.5B and 3B, Gemma 2B) now return usable clips instead of sometimes finishing with none. In our tests on a 20-minute video, Qwen 1.5B went from 1 clip to 7.
- If the AI's picks still can't be used, ClipShip picks clips from your transcript instead of stopping, and tells you a larger AI model usually picks better moments.
- Damaged or cut-off videos no longer fail. ClipShip uses every part it can read and tells you how much of the file was readable.
- Clear messages when an export can't be saved because the file is open in another app, the folder is protected by Windows, or the disk is full.
- Help & Support now sends your report straight to ClipShip support and shows your ticket number, instead of opening your email app.
Notes
- Your source video stays on your computer. A support report includes your app version, system details and recent activity, never your video, audio, transcript or API key.
- Pricing, subscriptions, referrals, device limits, and existing customer access are unchanged.
v1.0.42: Older PCs, smoother setup and caption exports
October 1, 2026
This update fixes setup getting stuck on some PCs, lets local AI run on older processors, and fixes several ways a captioned export could fail.
Changes
- Fixes setup failing over and over on some PCs. ClipShip now repairs its own files if they were left incomplete.
- Local AI now works on older processors and many Celeron and Pentium laptops, where it used to stop with an error.
- If the transcription model is missing, ClipShip offers to download it and continue, instead of stopping with an error.
- Fixes captioned exports failing with "FFmpeg not found" on some PCs.
- Clips with no spoken words now export without captions instead of failing.
- Long clips no longer stop while their captions are being finished.
- Fixes a crash when you reopen a project and go back to change the style.
- Videos with no speech now get a clear message instead of an error.
- If your monthly renewal didn't go through, Settings now lets you update your payment method and keep your current price.
Notes
- Your source video stays on your computer. Optional API mode sends transcript text directly to the AI provider you select.
- Pricing, subscriptions, referrals, device limits, and existing customer access are unchanged.
v1.0.41: Smoother restarts and missing AI model recovery
September 27, 2026
This update fixes two ways a run could stop early: starting again after leaving the processing screen, and a local AI model that was no longer on your computer.
Changes
- Fixes "A transcription is already running" when you leave the processing screen and start again. ClipShip now stops the earlier run and continues with the new one.
- If the local AI model you picked is missing, ClipShip now offers to download it right there and then runs your video again, instead of stopping with an error.
Notes
- Your source video stays on your computer. Optional API mode sends transcript text directly to the AI provider you select.
- Pricing, subscriptions, referrals, device limits, and existing customer access are unchanged.
v1.0.40: YouTube imports and local AI on more PCs
September 26, 2026
This update fixes the main reason YouTube, Twitch, and Kick imports were stopped as stuck, and fixes local AI failing to start on Windows PCs that were missing a Microsoft system file.
Changes
- Fixes online imports being stopped as stuck while they were still downloading. Slow connections now finish, and the import shows real download progress.
- Fixes local AI failing to start on some Windows PCs because a required Microsoft runtime file was not included with ClipShip.
- Explains clearly when a video has no audio track instead of failing transcription.
- If you chose GPU mode and your graphics driver fails, ClipShip now tells you to switch to Automatic so it can use the CPU.
- Closing an unsaved project no longer shows an error when Windows is still holding one of its files.
Notes
- Before release, this version was tested end to end on real videos: GPU and CPU-only processing, clip rendering, captioned export, and a slow YouTube download.
- Your source video stays on your computer. Optional API mode sends transcript text directly to the AI provider you select.
- Pricing, subscriptions, referrals, device limits, and existing customer access are unchanged.
v1.0.39: Local AI and import reliability fixes
September 25, 2026
This update fixes the causes behind several recurring processing failures, including a bug that stopped the Qwen 2.5 1.5B model from analyzing videos at all.
Changes
- Fixes Qwen 2.5 1.5B, the model recommended for lighter computers, which failed to start clip analysis since 1.0.36.
- Lets local AI analysis keep running on slower CPU-only computers while it is still making progress, instead of stopping at a fixed 10 or 20 minute limit.
- Frees the local AI model reliably before transcription starts, so a retry after a long analysis no longer fails immediately.
- Checks downloaded AI model files for damage and asks you to download a damaged model again instead of showing a generic load error.
- Retries once on the CPU when GPU transcription crashes in Automatic mode.
- Retries a YouTube, Twitch, or Kick import after 90 seconds of silence while reading the video details, instead of waiting 8 minutes.
- Stops a caption render only when it has genuinely stopped making progress, closes its background browser processes, and retries once with lower memory use.
- Shows a short, plain explanation when a clip export fails.
Notes
- Crash reports now include the technical cause of a failure with file paths, names, emails, links, clip titles, and transcript text removed, plus a random install ID that is not linked to your account. This lets us fix problems we previously could not see.
- Your source video stays on your computer. Optional API mode sends transcript text directly to the AI provider you select.
- Pricing, subscriptions, referrals, device limits, and existing customer access are unchanged.
v1.0.38: Clearer answers when processing fails
September 24, 2026
This update makes recurring processing failures easier to diagnose without sending video titles, file paths, or transcript text in crash reports. It also prevents a caption-render timeout from triggering another long wait.
Changes
- Separates known local AI startup and API-provider failures into clearer, privacy-safe support categories.
- Records the last safe stage of an online import or transcription failure so support can see where a job stopped.
- Keeps video titles, local paths, and raw provider output out of new import and export crash reports.
- Stops a timed-out caption renderer without repeating another ten-minute attempt.
Notes
- This release improves diagnosis and bounds one timeout. It does not claim that every underlying AI, network, transcription, or caption-render failure is fixed. If a job still fails on 1.0.38, please contact support with the app version and what you were doing.
- Your source video stays on your computer. Optional API mode sends transcript text directly to the AI provider you select.
- Pricing, subscriptions, referrals, device limits, and existing customer access are unchanged.
v1.0.37: More reliable first-time setup
September 23, 2026
This update tightens the checks that happen before ClipShip opens Create, so an incomplete API-key or transcription setup does not send you back to the hardware scan without a clear reason.
Changes
- Checks that a saved API key can be read back before marking API setup complete.
- Checks the required transcription model before opening Create and returns to its setup step if a download is still needed.
- Keeps your completed setup intact when secure key storage is temporarily unreadable, instead of silently restarting onboarding.
- Shows a retryable error if ClipShip cannot save your setup progress.
- Takes you to the missing setup step if hardware detection fails and you choose to continue.
Notes
- An API key covers clip analysis, not transcription. The transcription model may still need a one-time download.
- A failed download or unavailable AI provider can still interrupt setup or processing; the app will now make the setup failure clearer.
- Pricing, subscriptions, referrals, device limits, and existing customer access are unchanged.
v1.0.36: Clearer setup and steadier processing
September 22, 2026
This update fixes an API-key setup loop and local AI model-selection mistakes, and improves handling of longer AI analysis and online imports.
Changes
- Completes API-key setup without sending you back through local AI hardware setup when no local AI model is installed.
- Uses the local AI model you selected instead of silently loading a different installed model.
- Gives active API analysis enough time for a provider retry while retaining a safety limit for jobs that stop making progress.
- Keeps online import progress readable when a video title contains characters Windows cannot decode.
- Shows the transcription model download size when available and asks before downloading it, including in API mode.
- Preserves clearer, privacy-safe failure details for transcription and export support cases.
Notes
- An API key covers clip analysis, not the local transcription model. The transcription model may still need a one-time download.
- Network interruptions, unavailable source links, provider outages, and damaged video files can still cause a job to fail; this release does not claim to eliminate those external failures.
- Local source video stays on your computer. Optional API mode sends transcript text directly to the AI provider you select.
- Pricing, subscriptions, referrals, device limits, and existing customer access are unchanged.
v1.0.35: More reliable local processing and AI recovery
September 19, 2026
This update repairs the remaining Windows first-launch video runtime timing issue and makes ClipShip's local AI retry return structurally valid clip results.
Changes
- Rechecks ClipShip's bundled FFmpeg and FFprobe tools whenever video processing needs them, including after a Windows first-launch timing miss.
- Verifies Python, FFmpeg, and FFprobe through the installed Windows app before a release can be published.
- Uses a constrained JSON response for the local AI's second attempt when the first response is incomplete or malformed.
- Groups processing alerts by the actual stage and cause so support can distinguish import, transcription, AI, and render failures.
- Keeps expected private, unavailable, regional, rate-limited, and network link failures out of software-error alerts while preserving useful local guidance.
Notes
- Supported-link availability can still depend on the source platform, region, sign-in requirements, and network conditions.
- Local AI processing and source video files remain on the computer. Optional API mode still sends transcript text directly to the provider selected by the customer.
- Pricing, subscriptions, referrals, device limits, and existing customer access are unchanged.
v1.0.34: Reliable Windows processing setup
September 16, 2026
This update fixes a Windows first-launch timing issue that could incorrectly report that Python was missing even though ClipShip's local processing runtime was installed.
Changes
- Rechecks the installed ClipShip runtime whenever a processing command needs it instead of relying only on the first startup probe.
- Recovers automatically when Windows finishes making the bundled runtime available shortly after ClipShip opens.
- Shows clear repair guidance when Windows Security or an incomplete installation has actually removed a required runtime file.
- Groups related setup failures by their underlying cause so support alerts no longer split the same problem across model checking and downloading.
- Adds regression coverage for rediscovering the installed Python runtime after a delayed first startup probe.
Notes
- This is a Windows reliability fix. macOS is included at the same version so updates remain consistent across platforms.
- Local AI processing and source video files remain on the computer. Optional API mode still sends transcript text directly to the provider selected by the customer.
- Pricing, subscriptions, referrals, device limits, and existing customer access are unchanged.
v1.0.33: Reliable caption exports and clearer link-import errors
September 15, 2026
This update repairs a Windows export failure where the caption renderer could return an unfinished overlay file, and it makes supported-link import failures more useful without exposing private source details.
Changes
- Validates every rendered caption overlay as a real readable video before the final export begins.
- Waits for a successful render to finish writing its MOV metadata and retries once into a separate temporary file when validation still fails.
- Stops the export safely instead of passing an incomplete caption overlay into the final FFmpeg composite.
- Groups the same export cause into one Sentry issue across Clip Editor and Ship so repeated attempts do not create duplicate issue groups.
- Separates private, sign-in, regional, rate-limit, unavailable-source, and network failures for YouTube, Twitch, and Kick imports while keeping URLs, titles, and local paths out of Sentry.
- Extends the finished Windows installer test from a short sample to a 22.5-second captioned export matching the customer failure duration.
Notes
- The export recovery is bounded: ClipShip waits briefly for finalization and performs at most one fresh caption-render retry.
- Local AI processing and source video files remain on the computer. Optional API mode still sends transcript text directly to the provider selected by the customer.
- Pricing, subscriptions, referrals, device limits, and existing customer access are unchanged.
v1.0.32: More reliable AI analysis and model setup
September 15, 2026
This update repairs three failure paths seen in real customer diagnostics: incomplete Gemini responses, a local AI model that cannot load on the GPU, and an interrupted transcription-model setup.
Changes
- Requests schema-validated JSON from Gemini, provides a larger bounded response allowance, and uses minimal thinking for supported Gemini 3 models so clip results are less likely to end mid-response.
- Retries an invalid AI response once with a genuinely smaller one-clip response before showing clear provider guidance.
- Falls back to CPU once when Local AI is set to Auto and the model cannot load on the GPU because of a CUDA, allocation, or graphics-memory failure.
- Resumes a transcription-model download when the provider reports completion but required files are still missing.
- Separates provider-output, local-model, import, transcription setup, and export failures into stable privacy-safe diagnostic groups.
- Stops model-generated text from being copied into normal diagnostic logs or Sentry error titles.
Notes
- Explicit GPU mode still fails instead of silently changing devices, and damaged model files do not trigger an unrelated CPU fallback.
- Local AI processing and source video files remain on the computer. Optional API mode still sends transcript text directly to the provider selected by the customer.
- Pricing, subscriptions, referrals, device limits, and existing customer access are unchanged.
v1.0.31: Clearer recovery when Windows processing fails
September 14, 2026
This update preserves the real reason when transcription or rendering cannot continue, so ClipShip can show the right recovery guidance instead of collapsing different failures into one generic AI error.
Changes
- Carries privacy-safe reason codes from Windows transcription and clip rendering into the app's error and support flow.
- Distinguishes missing runtimes, permissions, storage, memory, GPU, source-file, caption-renderer, and timeout failures without exposing local paths or media content.
- Retries a temporarily unavailable caption-runtime startup once before asking the customer to repair the installation.
- Preserves both the beginning and end of long sanitized diagnostics so the useful failure detail is not cut off.
- Stops reporting a successfully recovered transcription as a new Sentry issue while retaining it as diagnostic context.
Notes
- The Windows release completed a physical import, NVIDIA CUDA transcription, Qwen 2.5 7B analysis, face-tracked render, review, and disk export test before publication.
- Local AI processing and source video files remain on the computer.
- Pricing, subscriptions, referrals, device limits, and existing customer access are unchanged.
v1.0.30: Short videos now get guidance, not a crash report
September 12, 2026
This follow-up makes the short-source check behave like normal guidance throughout the app, without opening the generic AI failure panel.
Changes
- Shows a dedicated Not enough spoken content message when the selected clip length cannot be created reliably.
- Provides one clear action to choose another source or a shorter clip length.
- Keeps expected short-source validation out of the crash-report panel, Sentry error reporting, and processing-failure analytics.
- Keeps unexpected AI, provider, and runtime failures on the existing diagnostic and support path.
Notes
- The spoken-material thresholds introduced in 1.0.29 are unchanged: 20 seconds for Short or Auto, 35 seconds for Medium, and 55 seconds for Long.
- Local AI processing and source video files remain on the computer.
- Pricing, subscriptions, referrals, device limits, and existing customer access are unchanged.
v1.0.29: Clear guidance for videos that are too short
September 12, 2026
This update stops local clip analysis early when a source does not contain enough spoken material for the selected clip length, instead of asking the AI to invent a result from too little context.
Changes
- Checks the amount of transcribed speech before local AI clip analysis begins.
- Shows a clear upload-longer-video message when there is not enough spoken material to make a reliable clip.
- Suggests the Short setting when a source is too brief for Medium clips, and Medium or Short when it is too brief for Long clips.
- Keeps this expected validation result out of Sentry while preserving genuine AI failure reporting.
Notes
- The minimum spoken material is 20 seconds for Short or Auto, 35 seconds for Medium, and 55 seconds for Long.
- Local AI processing and source video files remain on the computer.
- Pricing, subscriptions, referrals, device limits, and existing customer access are unchanged.
v1.0.28: Reliable transcription on clean Windows installs
September 12, 2026
This update fixes a Windows setup gap that could stop transcription before it started on computers without the Microsoft C++ runtime already installed.
Changes
- Bundled the required Microsoft C++ runtime directly with ClipShip's Windows transcription engine.
- Added an immediate startup check for the packaged transcription runtime before ClipShip begins extracting audio.
- Added clear reinstall guidance when Windows Security, an interrupted update, or a damaged install leaves the local transcription runtime incomplete.
- Strengthened the Windows release test to prove the finished installer loads ClipShip's own runtime files instead of relying on software already present on the build machine.
- Preserved structured, privacy-safe dependency failures in local diagnostics so support reports identify the correct recovery path.
Notes
- This repair is for Windows transcription startup; macOS remains supported in the same 1.0.28 release.
- Local AI processing and source video files remain on the computer.
- Pricing, subscriptions, referrals, device limits, and existing customer access are unchanged.
v1.0.27: Direct access to ClipShip support
September 10, 2026
This update connects the desktop Help & Support screen to ClipShip's live support page, so a customer can open the current help form without leaving the app to search for it.
Changes
- Added an Open support page action to the existing Help & Support screen.
- Open the verified https://clipship.co/support destination in the system browser.
- Kept the existing privacy-safe diagnostics copy action, FAQ link, and email option available in the same support screen.
Notes
- Opening the support page does not send a message; the customer reviews and submits the form themselves.
- ClipShip's diagnostic report continues to remove local paths and sensitive content before it can be copied.
- Pricing, subscriptions, referrals, device limits, and existing customer access are unchanged.
v1.0.26: More reliable local AI startup
September 5, 2026
This update fixes a failure after transcription when the local AI model needs more time to load, and gives clearer recovery guidance when the model cannot start.
Changes
- Wait for the selected model to finish loading before counting transcript tokens or starting clip analysis.
- Allow up to three minutes for model startup, with visible loading progress instead of a misleading title-writing message.
- Stop promptly when the model reports a permanent loading failure, with specific guidance for missing files, an incomplete runtime, insufficient memory, or unavailable GPU support.
- Keep failed or partially loaded models unavailable for requests, and preserve the failure category in sanitized support diagnostics.
- Added regression tests for delayed startup, failed loading, interrupted connections, invalid token counts, and the startup deadline on both Windows and Linux.
Notes
- Local AI still processes the source video and transcript on the computer. This update does not automatically switch processing to a cloud provider.
- A model that cannot load now has a clear recovery path; this does not remove the model's hardware and memory requirements.
- Pricing, subscriptions, referrals, device limits, and existing customer access are unchanged.
v1.0.25: Reliable local AI handoff to transcription
September 2, 2026
This update fixes a Windows timing issue that could ask for an app restart while ClipShip was releasing the local AI model before transcription.
Changes
- Replaced the fixed local AI shutdown delay with a bounded check that waits until the model server has actually released its listener.
- Added a safe fallback that force-stops only ClipShip's own tracked model process when graceful shutdown does not finish in time.
- Applied the same confirmed shutdown behavior when restarting an unhealthy local AI server and when stopping it from the app.
- Added regression coverage for immediate, delayed, and unsuccessful shutdown paths.
Notes
- The source video, transcript, and local AI processing remain on the computer.
- The shutdown path remains fail-closed if ClipShip cannot verify that the local AI listener was released.
- This release does not change pricing, Dodo products, subscriptions, referrals, device limits, billing, or existing customer access.
v1.0.24: Reliable caption runtime checks and recovery
September 1, 2026
This update detects an incomplete local caption-rendering runtime before export and gives a clear recovery path instead of failing after a long render attempt.
Changes
- Added an export preflight that verifies the bundled Python, FFmpeg, FFprobe, Node.js, HyperFrames, and caption browser files are present and can start.
- Added direct reinstall and Windows Security recovery guidance when an updater, interrupted install, or quarantine leaves one of those bundled components unavailable.
- Kept projects and settings outside the reinstall path, so repairing the application does not remove saved ClipShip work.
- Added a dedicated privacy-safe runtime-incomplete category to export diagnostics so this installation problem is no longer grouped with unknown export failures.
- Strengthened the Windows release gate to unpack the finished installer and complete a real captioned export using only the runtime contained inside it.
Notes
- The source video, transcript, and caption render remain on the computer.
- Diagnostics identify the missing runtime component without including local file paths or source content.
- This release does not change pricing, Dodo products, subscriptions, referrals, device limits, billing, or existing customer access.
v1.0.23: Reliable local AI context and easier support
August 31, 2026
This update prevents long or multilingual transcripts from exceeding the loaded local AI model's real context window and makes it easier to send useful diagnostics when a run fails.
Changes
- Counted each local AI prompt with the exact tokenizer used by the model that is currently loaded instead of relying on an approximate character count.
- Added model-aware transcript chunking that preserves room for the response and a focused retry, including long multilingual transcripts that use more tokens per character.
- Added recursive re-chunking when a prepared request is still too large, so ClipShip can reduce the request safely instead of sending an impossible prompt to the inference engine.
- Changed the local AI server to reject an oversized request before inference with a clear context-window error instead of allowing it to fail or stall inside the model runtime.
- Added one Help & Support screen with FAQs, app and operating-system details, privacy-safe diagnostics, copy controls, and an email action.
- Connected the footer Help button, Settings support action, native Report Bug menu, and crash recovery screen to the same support workflow.
Notes
- The source video, transcript, and AI analysis remain on the computer when Local AI is selected.
- Support diagnostics remove local paths and sensitive content, and the email opens in the customer's mail app for review before anything is sent.
- This release does not change pricing, Dodo products, subscriptions, referrals, device limits, billing, or existing customer access.
v1.0.22: Reliable long-video imports and processing
August 21, 2026
This update repairs the failure paths that could leave long YouTube videos downloading or processing indefinitely, especially on Windows.
Changes
- Changed YouTube imports to prefer reliable HLS streams before direct DASH formats that can fail with HTTP 403 responses.
- Passed ClipShip's bundled Node.js runtime to yt-dlp so JavaScript-based YouTube extraction does not depend on a separate system installation.
- Added an eight-minute inactivity watchdog and a two-hour absolute limit so a stalled online import ends with a clear recovery message instead of running overnight.
- Recovered a completed transcript when the native Windows transcription runtime fails only during GPU shutdown after already writing valid output.
- Removed stale transcript output before every new run so a previous partial result can never be mistaken for the current transcription.
- Removed the duplicate outer local-AI retry that could repeat an already timed-out analysis and make processing appear frozen for much longer.
- Added privacy-safe failure categories for online import, transcription, and analysis diagnostics without collecting video URLs, transcript text, or local file paths.
Notes
- The source video remains on the computer after an online import completes. Local AI keeps transcription and clip analysis on the computer.
- A stalled import now fails visibly rather than continuing without progress, so it can be retried or reported with useful diagnostics.
- This release does not change pricing, Dodo products, subscriptions, referrals, device limits, billing, or existing customer access.
v1.0.21: Safer processing recovery and clearer errors
August 18, 2026
This update stops stalled local AI analysis cleanly, recovers more usable clips from imperfect AI timestamps, and gives clearer storage and export errors.
Changes
- Stopped a local AI request after its 10-minute inference limit instead of retrying the same stalled request and making the run appear frozen.
- Added one focused retry with stricter timestamp rules when every AI-selected clip fails validation.
- Added a safe relaxed-validation recovery for imperfect but still usable clip segments instead of dropping the entire result immediately.
- Verified that the selected storage folder is writable before starting multi-gigabyte model downloads, with clear permission and disk-space guidance when it is not.
- Hardened Copy crash log so clipboard focus failures fall back safely and show a useful retry message instead of failing silently.
- Added privacy-safe analysis and export diagnostics that include the failed stage, model, device, and error category while removing local file paths.
Notes
- The source video, transcript, and AI analysis remain on the computer when Local AI is selected.
- The added diagnostics do not include source video contents, transcript contents, model contents, or local filesystem paths.
- This release does not change pricing, Dodo products, subscriptions, referrals, device limits, billing, or existing customer access.
v1.0.20: Reliable language selection and AI clip recovery
August 17, 2026
This update keeps the selected spoken language attached to the project and recovers when a local AI model initially selects no clips.
Changes
- Kept the spoken-language selection stable when moving between clip setup and processing instead of allowing the screen to reset it silently.
- Added a processing-stage language indicator so the selected transcription language is visible before captions are generated.
- Added a fail-closed check that stops the run if the transcription engine reports a different forced language than the one requested.
- Changed valid empty AI results so they are recognized as no clip selection instead of being reported as malformed JSON.
- Added a focused second AI pass that selects the strongest available moment when the first pass returns no clips.
- Added a clear recovery message when a video still produces no usable clips after both analysis passes.
Notes
- Genre changes what kinds of moments ClipShip prioritizes. It does not change the spoken language used for transcription.
- The source video, transcript, and AI analysis remain on the computer when Local AI is selected.
- This release does not change pricing, Dodo products, subscriptions, referrals, device limits, billing, or existing customer access.
v1.0.19: Safer 4K transcription and clearer clip setup
August 16, 2026
This update prevents local AI and transcription from competing for graphics memory, makes 4K sources safer to process, and adds clearer controls for the clips ClipShip should create.
Changes
- Changed transcription to work from a temporary 16 kHz mono audio file instead of decoding the original video container inside the transcription engine.
- Added an exclusive transcription lock so the local language model is fully stopped before transcription starts and cannot reload until transcription finishes.
- Added automatic CPU fallback when available NVIDIA graphics memory is too low for reliable transcription, with a clear warning when forced GPU mode cannot run safely.
- Added a spoken-language selector so the transcription engine can use the correct language hint instead of relying only on automatic detection.
- Added a real Smart Mix or Keep Full Frame choice to clip setup and carried that choice through analysis, review, and rendering.
- Added cleanup for temporary transcription audio even when a run is cancelled or fails.
Notes
- The source video remains on the computer. The temporary transcription audio is local and is removed after the run.
- On Apple Silicon, ClipShip continues to use MLX Whisper and Metal. On Windows, GPU transcription remains available when the selected mode and available graphics memory can support it safely.
- This release does not change pricing, Dodo products, subscriptions, referrals, device limits, billing, or existing customer access.
v1.0.18: Reliable transcription engine setup
August 15, 2026
This update makes the first transcription-engine download resilient to interrupted connections and shows a useful error when setup cannot continue.
Changes
- Added bounded retries and resumable downloads for temporary connection failures during transcription-engine setup.
- Added a complete-file verification step so onboarding cannot continue with a partial or damaged transcription model.
- Added clear inline recovery messages for network, TLS, storage, server, and bundled-runtime failures instead of returning to an unchanged setup screen.
- Added privacy-safe failure categories to crash reporting so setup problems can be diagnosed without collecting file paths, model contents, or personal data.
- Added a bundled-runtime check before downloading so an incomplete installation is identified immediately with the correct reinstall guidance.
Notes
- Existing complete transcription models remain in place and are not downloaded again.
- Windows continues to use Faster-Whisper and Apple Silicon Macs continue to use MLX Whisper; the recovery behavior is shared across both platforms.
- This release does not change pricing, Dodo products, subscriptions, referrals, device limits, billing, or existing customer access.
v1.0.17: Offer codes and creator attribution
August 14, 2026
This update adds a durable offer-code fallback for creator referrals and enforces ClipShip's founder lifetime allocation at checkout.
Changes
- Added one optional Offer code field during account setup for referral, creator-partner, and future promotional codes.
- Added the same optional Offer code field before Pro checkout so attribution can be recovered even when a download and later signup happen on different networks.
- Added a 14-day Pro trial for valid creator-partner offers while keeping the normal ClipShip trial at seven days.
- Added automatic recent-link detection where the match is unambiguous, with the visible offer code as the reliable fallback.
- Added automatic-link and offer-code attribution counts to the creator dashboard alongside verified clicks, accounts, conversions, and commission entries.
- Enforced the $99 founder lifetime allocation atomically for the first 100 eligible lifetime purchases, then switched new lifetime checkouts to the $149 standard product.
Notes
- The installer is identical for every customer and contains no buyer-specific or creator-specific data.
- Creating another account on the same computer does not restart or extend a used trial.
- Existing subscriptions, existing lifetime licenses, device limits, signed-webhook activation, and customer access are unchanged.
v1.0.16: Safer Windows updates and AI provider repairs
August 13, 2026
This update prevents incomplete Windows reinstalls and repairs the provider failures found in recent production diagnostics.
Changes
- Changed the Windows installer to detect and close only stale ClipShip runtime processes before replacing bundled files.
- Removed the installer Ignore path that could leave Python DLLs missing and make ClipShip fail to open after an update.
- Added a release-blocking import check for the exact bundled transcription, local AI, SSL, database, vision, and media runtime components.
- Reduced Groq request sizes for lower free-tier token limits and added bounded retry guidance for rate-limited requests.
- Made local AI failures report the actual model or runtime loading problem instead of incorrectly saying that the server is unreachable.
- Removed Gemini's unbundled Python SDK dependency, replaced retired Gemini 2.0 defaults, and migrated saved retired defaults to Gemini 3.1 Flash-Lite.
Notes
- The locked-DLL installer failure affected Windows reinstalls and updates only. The macOS DMG does not use NSIS or Windows Python DLLs.
- The shared Groq, Gemini, and local AI error repairs are included in both the Windows and macOS builds.
- This release does not change the current $99 lifetime charge, Dodo products, subscriptions, referrals, device limits, billing, or existing customer access.
v1.0.15: Recovery and CPU stability
August 11, 2026
This update fixes two Windows failures found in production diagnostics and makes error reporting more accurate.
Changes
- Fixed recovered-session cleanup when Windows briefly keeps a project file open after a preview or process exits.
- Changed the Discard action to show a useful retry message instead of producing an unhandled app error if a file remains locked.
- Isolated explicit CPU transcription from optional NVIDIA CUDA runtime files that could crash the CPU process on some Windows systems.
- Stopped provider rate limits and quota responses from being reported as ClipShip crashes while keeping the error visible in the app.
- Completed privacy-safe export milestone tracking for people who opted into anonymous usage statistics.
Notes
- The macOS build is updated to the same version so both signed updater channels remain aligned. Its Metal and MLX processing architecture is unchanged.
- Anonymous product analytics still excludes account identity, file names, paths, source URLs, transcripts, prompts, API keys, clip text, and exported media.
- This release does not change pricing, products, subscriptions, referrals, device limits, billing, or existing customer access.
v1.0.14: Clearer Free and Pro feature boundaries
August 7, 2026
This update keeps ClipShip's original Free workflow intact while moving the newer workflow controls and publishing tools into Pro.
Changes
- Added clear Pro locks and upgrade actions for manual target clip counts, on-video headlines, and generated posting copy.
- Kept automatic clip finding, built-in caption styles, editing and reframing, and 720p watermarked exports available on Free.
- Added native entitlement checks so paid processing and export features cannot be unlocked by changing the interface alone.
- Made Free processing fall back to automatic clip counts and made Free exports omit previously saved paid headline overlays.
- Cleared paid export caches when an account's entitlement changes so an expired trial cannot reuse a prior Pro render.
Notes
- The seven-day trial still includes every Pro feature.
- Existing Pro subscribers and Lifetime Pro owners keep the same features and billing terms.
- This release does not change prices, product IDs, subscriptions, referrals, device limits, or existing customer access.
v1.0.13: Give 20%, get 20% referrals
August 7, 2026
This update adds a referral program for the ClipShip lifetime license, with a reward for both the buyer and the person who referred them.
Changes
- Added a personal referral code and shareable link in Settings.
- Added 20% off a first Lifetime Pro purchase for an eligible referred buyer.
- Added one 20% Lifetime Pro discount for referrers on Free, Trial, or Monthly after the referred purchase clears its refund window.
- Added banked permanent device slots for later rewards, which activate automatically after the referrer owns Lifetime Pro.
- Added referral status, successful-referral count, remaining rewards, and available-discount status in Settings.
- Added repeat-device signup risk records for internal review without blocking legitimate account access or purchases.
Notes
- Referral discounts apply only to the $99 lifetime license. The $5 monthly plan and extra-slot purchases are not discounted.
- Rewards become available after the referred buyer's seven-day refund window and are limited to three successful referrals per account: at most one Lifetime discount, followed by device slots.
- Self-referrals, shared-device referrals, discount stacking, refunded payments, disputes, and chargebacks do not qualify.
- The signed Dodo webhook remains the only authority for activating Pro access and qualifying referral rewards.
v1.0.12: ClipShip for macOS and safer offline access
August 6, 2026
This update brings ClipShip to Apple Silicon Macs, strengthens license verification, and fixes the model, export, project, and progress issues found during Mac testing.
Changes
- Added the first public Apple Silicon macOS build for macOS 15 and newer, distributed as a signed and Apple-notarized drag-to-Applications DMG.
- Added native Mac transcription through MLX and local clip analysis through llama.cpp Metal, with unified-memory checks before processing.
- Added signed automatic updates for macOS through the same ClipShip update channel used by Windows.
- Made the Whisper onboarding download use ClipShip's managed model cache with visible checking, downloading, verifying, ready, and error states.
- Fixed custom caption styles in final Mac exports, Command-V in text and API-key fields, stale pipeline reuse after changing AI mode, and unsaved-project prompts.
- Added truthful transcription, analysis, render, export, and YouTube-upload stage feedback, plus consistent bulk-export counts.
- Improved cancellation and failure reporting so intentional stops are not reported as crashes and provider-key errors remain understandable.
- Added editable spoken openings, short on-video headlines, and post captions with Direct, Curiosity, and Bold alternatives grounded in the transcript.
- Added optional privacy-safe product milestones that exclude account identity, file names, video content, transcripts, captions, prompts, and API keys.
- Changed paid Pro and Pro trial access to require a successful online license check at least once every 24 hours, including while the app remains open.
Notes
- The free trial still lasts seven calendar days. The 24-hour rule is the maximum time Pro or trial features can continue without a successful online verification.
- When the seven-day trial ends, ClipShip continues on the Free plan after the next successful license check. Free accounts retain a seven-day offline window.
- The Mac release supports Apple Silicon only. Intel Macs are not supported in this release.
- The first model download and online features still require internet. Local AI mode continues to process the source video, transcript, captions, and exports on the device.
- Windows receives the shared workflow, diagnostics, and licensing fixes. Its existing CUDA and CPU processing architecture is unchanged.
v1.0.11: More clips and editable headlines
July 27, 2026
This update gives you direct control over how many clips ClipShip looks for and adds export-ready headlines that stay editable per clip.
Changes
- Added Auto, up to 5, up to 10, up to 15, and up to 20 clip targets before processing.
- Made Auto scale the target to the source video's duration instead of using one fixed count for every video.
- Improved larger runs by analyzing the transcript in smaller reliable batches, retrying malformed AI responses per batch, and balancing picks across the full video.
- Added an option to start every generated clip with an AI-generated headline based on its clip title.
- Added independent headline text, visibility, style, position, size, text color, and accent color controls for every clip.
- Added Bold, Clean, and Boxed headline styles plus an Apply appearance to all clips action that keeps each clip's own text and visibility setting.
- Added headline rendering to final exports with matching live preview placement.
Notes
- Clip targets are maximums. ClipShip can return fewer when the source does not contain enough distinct usable moments.
- Choosing more clips increases local AI analysis and rendering time, especially on CPU-only systems.
- Headlines are stored separately from captions, so they can be used even when animated captions are turned off.
v1.0.10: Twitch and Kick video imports
July 25, 2026
This update expands link imports beyond YouTube while making the supported video sources and limitations explicit.
Changes
- Added imports for Twitch VODs and clips.
- Added imports for Kick clips.
- Replaced the YouTube-only link option with one video-link importer that identifies the supported platform automatically.
- Added clear YouTube, Twitch, and Kick source labels beside the link field.
- Rejects live streams, unsupported websites, lookalike domains, and non-video profile or channel pages before downloading.
- Updated and pinned the bundled video downloader to the tested 2026.07.04 release with checksum verification.
Notes
- Supported links are YouTube videos and Shorts, Twitch VODs and clips, and Kick clips.
- Kick VOD imports are temporarily unavailable because Kick's current VOD metadata endpoint is not working with the downloader.
- Only import videos you own or have permission to use.
v1.0.9: Account and checkout protection
July 24, 2026
This update closes an account-state gap that could let an incomplete signup enter ClipShip before the product account was created.
Changes
- Allows only verified trial, free, or active license states to enter the application.
- Blocks unknown or incomplete account states instead of treating them as usable licenses.
- Prevents the Pro plan chooser from opening until the ClipShip account and license have been verified.
- Returns incomplete accounts to signup if checkout discovers that the product account is missing.
- Fixes the required-update action so it opens the built-in updater even when that version was previously skipped.
Notes
- The checkout backend already prevented incomplete accounts from being charged. This release also prevents those accounts from reaching checkout in the first place.
- No payment, subscription, or existing license data is changed by installing this update.
v1.0.8: Safer account setup and upgrades
July 22, 2026
This update makes account and license state clearer before an upgrade, preventing a payment from being started before ClipShip can attach it to the correct account.
Changes
- Returns incomplete accounts to the signup flow instead of leaving Settings on an unknown plan status.
- Shows the account setup error directly in Settings so the next required step is clear.
- Hides upgrade actions until account setup and license verification are complete.
- Prevents checkout from starting when the signed-in account does not yet exist in ClipShip's license database.
- Reconciles older accounts safely when Firebase identity details have changed but the verified email uniquely matches the existing ClipShip account.
Notes
- No payment or license data is changed by installing this update.
v1.0.7: Monthly Pro plan and billing choices
July 22, 2026
This update adds a $5 founding monthly plan alongside the existing $99 lifetime license, with the same Pro features on either plan.
Changes
- Added a $5-per-month founding Pro plan for the first 50 paid subscribers. Founding subscribers keep that price while their subscription remains continuously active.
- Kept the $99 one-time lifetime license as a separate option with no recurring charge.
- Changed the in-app Upgrade action to show a monthly-versus-lifetime plan chooser before opening checkout.
- Added billing details in Settings so monthly subscribers can see their plan state and open the secure billing portal to manage or cancel their subscription.
- Added a switch-to-lifetime option for monthly subscribers. The monthly subscription is cancelled after the lifetime payment succeeds.
- Improved the post-checkout license refresh so newly activated Pro access and additional device slots appear without relying on stale app state.
Notes
- Monthly and lifetime licenses unlock the same Pro features: watermark-free 1080p exports and custom caption styles, with one active device included.
- Existing lifetime licenses are unchanged. No current lifetime customer is moved to a subscription.
- The $5 founding rate is retained only while the subscription remains active. Cancelling and joining again later may mean paying the then-current monthly price.
- ClipShip still processes videos locally on your PC. The new subscription option changes billing, not where your videos or AI processing run.
v1.0.6: AI analysis recovery fix
July 7, 2026
This update improves the AI clip-selection step when a local model returns malformed or overlapping clip candidates.
Changes
- Added recovery for overlapping AI-selected segments inside the same clip instead of failing the whole run.
- Added a compact JSON retry when the AI returns malformed or truncated clip data.
- Improved validation so ClipShip can keep a usable imperfect clip instead of returning zero clips after analysis.
- Fixed Sentry diagnostics so local AI errors are tagged as local instead of being mislabeled as Gemini.
Notes
- This fix targets analysis failures where the app said AI returned clips but all were dropped during validation.
- The app still keeps strict validation first. The recovery path only runs when strict validation would otherwise leave the user with no clips.
- If AI analysis still fails, email [email protected] with the crash report so the exact transcript/model behavior can be inspected.
v1.0.5: Caption export fix
June 6, 2026
This update fixes exports that failed after clip processing completed because the caption renderer was missing on clean Windows installs.
Changes
- Bundled the local caption-rendering runtime into the Windows installer instead of depending on a globally installed HyperFrames command.
- Bundled the headless browser runtime used for animated caption overlays during export.
- Changed the export path so ClipShip prefers its bundled renderer first and only falls back to global commands on development machines.
- Improved the missing-runtime error message so future packaging issues are easier to diagnose.
Notes
- This is the fix for exports that failed with HyperFrames CLI not found after processing clips successfully.
- No manual npm install or command-line setup is required on your PC.
- The installer is larger in this version because the caption export renderer now ships with ClipShip.
v1.0.4: RTX 50-series GPU compatibility build
June 6, 2026
This update adds a separate Blackwell CUDA runtime path for RTX 50-series GPUs and improves GPU runtime diagnostics.
Changes
- Added RTX 50-series / Blackwell GPU detection so ClipShip downloads a newer CUDA runtime pack instead of reusing the older legacy CUDA stack.
- Kept RTX 20/30/40-series cards on the existing legacy CUDA runtime path so older supported GPUs do not receive an unnecessary runtime change.
- Separated the Blackwell CUDA files into their own local folder to avoid stale DLLs from the legacy runtime mixing with the newer runtime.
- Updated transcription and local AI startup logs to include GPU name, runtime flavor, and runtime version.
- Improved runtime-pack handling so downloadable CUDA zips stay outside the Windows installer bundle.
Notes
- RTX 50-series support depends on the installed NVIDIA driver and the external CUDA pack downloaded from ClipShip's CDN.
- If you are testing on an RTX 50-series card, open Settings and set Local processing device to GPU before starting the run.
- If it still fails, email [email protected] with the crash report so the exact GPU/runtime error can be inspected.
v1.0.3: Transcription and GPU mode fix
June 4, 2026
This update fixes clean-install transcription setup and makes the GPU/CPU setting apply consistently to local processing.
Changes
- Fixed the transcription model mismatch so the app transcribes with the same Whisper model that onboarding downloads and verifies.
- Fixed Whisper model checks and downloads so they respect the user's selected model storage folder.
- Fixed the Settings local-model download button so it downloads local clip-finding AI models instead of calling the transcription-engine downloader.
- Fixed the processing pipeline so the selected local processing device is passed to Whisper transcription instead of always using Auto.
- Changed forced GPU mode so transcription fails with a clear CUDA error instead of silently falling back to CPU.
Notes
- If ClipShip still says the transcription model is missing after updating, open Settings and download the transcription engine once.
- Auto mode can still choose the safest available device. GPU mode now means GPU is required.
v1.0.1: Stability update
June 4, 2026
This update focuses on the two issues reported after launch: local AI device mode and the unsaved-project close prompt.
Changes
- Fixed Local AI processing mode so choosing CPU actually forces the clip-selection AI to run on CPU.
- Kept transcription device selection separate, so Whisper can still use GPU automatically when it is safe while CPU mode controls the local AI clip finder.
- Added stricter GPU mode behavior. If GPU mode is selected and CUDA is not really being used, ClipShip fails clearly instead of silently falling back.
- Reduced Settings flicker by preloading hardware and CUDA status when the app starts.
- Improved the unsaved-project close prompt so the app cleanup waits until the window is really closing.
- Cleaned up local AI error messages so failed model/server startup is easier to understand.
Notes
- First-time setup still needs internet for the bundled model, transcription runtime, and optional CUDA downloads.
- Windows SmartScreen may still warn on install because ClipShip does not yet have an EV code-signing certificate.
v1.0.0: Public launch
May 21, 2026
The first public Windows release of ClipShip, a local AI clip generator for turning long talking-head videos into short-form clips.
Changes
- Local video processing for transcription, clip selection, face-tracking vertical crop, captions, and export.
- Free tier with 720p exports and a Pro tier with 1080p exports, no watermark, and custom caption styles.
- One-time Pro license with update support instead of a monthly subscription.
- Auto-updater support for future Windows releases.
Notes
- ClipShip is a public v1 desktop app. Bug reports and feature requests go to [email protected].
Latest download
New installs can always choose the latest Windows or Apple Silicon macOS installer from the ClipShip homepage.