Hardcore Reviews
Hardcore Reviews

Burn-In or Soft? Picking a Video Subtitle Remover

A comparison of eight options - CapCut, Vrew, Bcut, Premiere Pro, whisperX, OpenAI Whisper, VSR, and Subtitle Edit. The core framing: hard (burned-in) subtitles cannot be removed losslessly and can only be erased or covered, always lossy; soft (independent track) subtitles can be stripped losslessly. Includes a comparison table and scenario-based selection.

Published November 1, 20269 min read
<!-- video-subtitle-removal-tools-comparison-review | review | Burn-In or Soft? Picking a Video Subtitle Remover -->

Many people ask whether there is a tool that removes subtitles from a video with one click. The honest answer is not a simple yes or no. It depends first on which kind of subtitle you are dealing with. This article is about how to choose a subtitle-removing tool. It does not dig into the full editing workflow of every tool. We keep our focus on one question only: can the subtitle be removed, how is it removed, and what is the cost. Set the technical framing first, and the comparison that follows will not drift.

Why does this deserve a dedicated comparison? Because the need to remove subtitles is surprisingly varied. Some want to reuse someone else clip but hate the original captions. Some want to swap the language and re-translate. Some want to clean auto-generated captions off a screen recording. Others just want to wipe a logo-like hard subtitle. The right tool changes completely with the scenario. Pick wrong and you either fail to remove it or needlessly lose quality.

Burn-in vs. soft: know the two types

Subtitles exist in two technically distinct forms, and every difference in tool capability comes from this split.

Hard subtitles, also called burned-in or hardcoded subtitles, are baked into the video pixels. They are part of the image, not a separate switchable stream. Because of that, hard subtitles cannot be removed losslessly. Any tool can only erase them with AI inpainting, cover them with an overlay, or crop the frame to cut them out. All of these alter the pixels, so the picture quality necessarily drops. On complex or moving backgrounds, ghosting and color patches are common.

Soft subtitles, also called independent subtitle tracks or softcoded subtitles, live in a separate track or an external file such as SRT, ASS, or VTT. They are detached from the picture. You can turn off or delete that track, or throw away the external file, with zero loss to image quality.

One sentence to remember: only soft subtitles can be removed losslessly. Hard subtitles can only be erased or covered by any tool, and that is always lossy. The first step in choosing a tool is to identify which kind you have.

How do you tell which kind you have? Try turning the subtitle off in a player, or look for a standalone .srt, .ass, or .vtt next to the video file. If you can switch it off, or a separate subtitle file exists, it is soft. If the caption is glued to the picture no matter what you do, and no separate file exists, it is hard. Get this wrong and everything downstream is wasted.

Side-by-side comparison

The table below is a feature comparison, not a benchmark. The real capability gaps are clear at a glance.

ToolHard subtitle handlingSoft subtitle handlingLanguagesPricePrivacy
CapCutAI erase / crop / cover (lossy)delete track losslessmany (no official count)free + Pro $9.99/mo, $74.99/yrweb / AI to cloud
Vrewno (no image repair)delete its own track100+see official sitecloud
Bcutsmart erase (lossy)delete track losslessmainly Chinesefreelocal app
Adobe Premiere Procontent-aware fill / mask manual (lossy)delete caption track27 / 40+ (two figures)subscription (~$22.99/mo)local + CC cloud
whisperXnogenerate SRT/VTTmany (needs align model)free open sourcelocal
OpenAI Whispernogenerate SRT/VTT + ffmpeg strip99free open sourcelocal
VSRinpaint fill (lossy, hard only)n/aneeds GPU modelfree open sourcelocal
Subtitle EditOCR to text only, no eraseadd/remove track lossless32+ UI languagesfree open sourcelocal

Prices are subject to official sites. For example, CapCut Pro US baseline is about $9.99 per month and $74.99 per year; others follow real-time official pricing. Where official language counts are not unified, we say so rather than invent a number.

Tool-by-tool notes

CapCut

CapCut and its Chinese-language build, Jianying, share the same engine. For soft subtitles, if the caption is still an editable layer or track, such as auto-captions inside the project, you can select and delete it on the timeline with no quality loss. For hard subtitles, they cannot delete directly. You must crop the frame, stack a cover layer, blur, or use the AI erase tool to paint over the background. All of these are lossy repairs. Basic editing and auto-captions are free; advanced AI erase has free-tier limits and needs a subscription beyond that, with regional pricing that runs cheaper in some markets. On privacy, the web version and some AI features upload to the cloud, which matters for sensitive footage.

Vrew

Vrew positions itself as an AI subtitle video editor built around editing video like a document. You upload a video, it transcribes speech automatically, and you add or remove clips and subtitles by editing the transcript text. What it deletes is its own generated subtitle track, a soft subtitle. Its public feature set does not include image repair or object removal, so it cannot erase burned-in hard subtitles. If you want to cut a video using subtitles, it is a smooth fit. If you came for hard-subtitle removal, it will not help. Transcription supports 100+ languages and runs in the cloud; the free tier exports with a watermark, and pricing follows the official site.

Bcut

Bcut is Bilibili official free editor, with one-click speech-to-text captions, smart erase, and frame cleanup, aimed at the Bilibili creator ecosystem, available on Windows, macOS, and mobile. For hard subtitles it offers smart erase and AI removal, both lossy, plus mosaic, blur, or crop covering. Soft subtitles can be removed by deleting the track. It is completely free, ad-free, and watermark-free, runs as a local app, and is privacy friendly. The weak point is repair quality on busy moving backgrounds, where a desktop-grade tool still does better.

Adobe Premiere Pro

Premiere Pro has no native one-click subtitle remover. For soft subtitles, Text-Based Editing imports an automatic transcript and lets you cut clips by editing text; its own caption track can be deleted, which is lossless. For hard subtitles you must use Content-Aware Fill, masks, pixel replacement, or object masks to repair region by region, which is lossy and tedious, or add a third-party plugin. It suits professionals who demand high precision and can spend the time. Subscription is about $22.99 per month billed annually; the Creative Cloud Pro bundle with 20+ apps has a promotional rate, and a 7-day trial is offered. Language counts vary by source, 27 versus 40+, so check the official site.

whisperX / OpenAI Whisper

Both are open-source local transcription tools, not subtitle removers but subtitle makers. OpenAI Whisper supports 99 languages and outputs TXT, VTT, SRT, and more, and can use ffmpeg to strip or keep soft subtitle tracks in a video, losslessly. whisperX adds forced alignment on top of Whisper for word-level timestamps and speaker diarization, outputting SRT, VTT, and JSON. Its models range from tiny to turbo, needing roughly 1 to 10 GB of VRAM, with default alignment for English, French, German, Spanish, Italian, and more. They only generate soft subtitles and cannot erase burned-in hard subtitles; for that you need a separate image-repair tool. They run fully locally, audio never leaves the machine, and privacy is strong.

VSR (open-source add-on)

VSR, short for video-subtitle-remover, is an Apache-2.0 open-source, offline local AI tool that removes burned-in subtitles and text watermarks through inpaint filling. It handles hard subtitles only; soft tracks do not apply. Most builds need an Nvidia GPU, run without uploading to the cloud, and are privacy strong; a SourceForge mirror is also available. It is one of the top free open-source choices for hard-subtitle removal, ideal for users willing to run a model locally and keep footage on their own machine.

Subtitle Edit (open-source add-on)

Subtitle Edit is a free open-source subtitle editor supporting 300+ format conversions. It can extract, add, or remove embedded soft subtitle tracks from MKV or MP4 with lossless stripping, ships local Whisper generation, and has a 32+ language interface at nikse.dk. For hard subtitles it can only OCR them into text and cannot erase the picture. It is the go-to tool for managing, stripping, and translating soft subtitles, and for pulling a video soft track out for further editing.

How to choose by scenario

  • Want lossless removal: first confirm it is a soft subtitle. If so, use Subtitle Edit, ffmpeg, or simply delete the track, all lossless.
  • Want to erase a hard subtitle: pick an erasing tool such as VSR (free local), CapCut AI erase, Bcut smart erase, or Premiere Pro Content-Aware Fill. Accept the loss.
  • Want to cut video through subtitles: a text-driven editor like Vrew is the smoothest.
  • Want to make your own subtitle track: generate SRT/VTT locally with Whisper or whisperX.
  • Privacy sensitive, no upload: prefer local open source, VSR, Subtitle Edit, Whisper, whisperX.

To confirm and strip a soft track without touching the picture, one command does it:

text
ffmpeg -i input.mp4 -c copy -sn output.mp4

Remember: hard subtitles are always lossy, soft subtitles alone are lossless. Identify the type first, then pick the tool.

One practical tip ties it together: always check the subtitle type before opening any app. If you only ever receive files with a sidecar .srt, the whole job is a one-line ffmpeg command and no editing suite is needed. If the text is burned into the frame, accept upfront that the result will carry repair artifacts, and judge tools by how clean their inpaint looks on your specific background rather than by marketing claims.

Three selection misconceptions

Misconception one: believing some tool can remove hard subtitles losslessly. None exists. Hard subtitles are fused into pixels; any erasing is repair, and picture information is always lost. When a tool says lossless, it usually means the resolution is unchanged, not that quality is preserved.

Misconception two: assuming every subtitle remover can handle hard subtitles. Tools like Whisper, whisperX, and Subtitle Edit either only make soft subtitles or only read text; they cannot erase the picture at all. Before buying, check whether the tool is an eraser or a stripper.

Misconception three: assuming the most expensive tool is the best. For soft subtitles, free open-source Subtitle Edit is enough. For hard subtitles, free open-source VSR holds up well. Subscriptions are priced for the whole editing suite, not necessarily for this one task.

FAQ

Q1: Can hard subtitles be removed losslessly?

A1: No. Hard subtitles are baked into the pixels. Any tool can only erase, cover, or crop them, and all of those alter pixels, so they are lossy. A tool that claims lossless resolution means the output resolution stays the same, not that picture information is preserved.

Q2: Is CapCut enough?

A2: It depends. For soft subtitles, deleting the track in CapCut is a lossless fix. For hard subtitles, AI erase or covering handles most daily needs, but busy moving backgrounds leave traces and quality drops. For high-precision professional work, move to Premiere Pro or VSR.

Q3: Any good open-source options?

A3: Yes. For hard-subtitle removal, choose VSR with local inpaint erasing. For managing and stripping soft subtitles, choose Subtitle Edit. To generate subtitle tracks locally, choose Whisper or whisperX. All three are free, run locally, and are privacy strong.

Q4: How should I think about privacy?

A4: If you do not want footage uploaded, avoid cloud tools such as the CapCut web version, Vrew, and CapCut cloud AI. Local apps like Bcut and the Premiere desktop build, plus local open-source tools VSR, Subtitle Edit, Whisper, and whisperX, keep material on your machine and are safer for privacy.

Q5: How do I process in batch?

A5: Consumer cloud tools mostly work per project, with limited batch ability. Real batch work uses scripts: Whisper, whisperX, and VSR all support command-line batch transcription or erasing, and Subtitle Edit handles batch subtitle-file operations. At scale, go scripted open source.

Related reading

This article is AI-assisted and human-edited. Last updated: 2026-11-01

FAQ

Can hard subtitles be removed losslessly?
No. Hard subtitles are baked into the pixels. Any tool can only erase, cover, or crop them, and all of those alter pixels, so they are lossy. A tool that claims lossless resolution means the output resolution stays the same, not that picture information is preserved.
Is CapCut enough?
It depends. For soft subtitles, deleting the track in CapCut is a lossless fix. For hard subtitles, AI erase or covering handles most daily needs, but busy moving backgrounds leave traces and quality drops. For high-precision professional work, move to Premiere Pro or VSR.
Any good open-source options?
Yes. For hard-subtitle removal, choose VSR with local inpaint erasing. For managing and stripping soft subtitles, choose Subtitle Edit. To generate subtitle tracks locally, choose Whisper or whisperX. All three are free, run locally, and are privacy strong.
How should I think about privacy?
If you do not want footage uploaded, avoid cloud tools such as the CapCut web version, Vrew, and CapCut cloud AI. Local apps like Bcut and the Premiere desktop build, plus local open-source tools VSR, Subtitle Edit, Whisper, and whisperX, keep material on your machine and are safer for privacy.
How do I process in batch?
Consumer cloud tools mostly work per project, with limited batch ability. Real batch work uses scripts: Whisper, whisperX, and VSR all support command-line batch transcription or erasing, and Subtitle Edit handles batch subtitle-file operations. At scale, go scripted open source.

Related

Hardcore Reviews

KV Cache Gap of 39x: Four Open-Weight Agent Flagships Compared

Four open-weight flagships compared: MiniMax M3.1 Flash (428B/23B active), DeepSeek V4.1 Flash 1017, GLM-5.3, and Kimi K3 - all four appear in MiniMax's official model-card comparison table, so calibers align naturally, with two caveats stated upfront: the whole table was run by MiniMax alone (the card admits it), and each benchmark used a different harness (Terminal-Bench 4.0 on Codex 0.149.1, DeepSWE on mini-SWE-agent at 512K, CyberGym on Claude Code); every M3.1 Flash score is vendor-reported with no third-party replication. The only third-party ruler is the AA Intelligence Index: DeepSeek 1017 at 45, M3.1 Flash 42, M3 at 29 (GLM-5.3's 59 is a different index in a different month, directional only). The real efficiency gap sits in KV cache: DeepSeek 1017 officially uses 890 bytes per token against M3.1 Flash's 34,560 (vLLM engineering figure) - a roughly 39x memory difference for self-hosted long context. There is no official M3-to-M3.1 benchmark table, only three claims (near-3x speed, 200+ tok/s, 70 percent lower agentic cost with no published method), a price cut (0.30/1.20 to 0.15/0.60 dollars), and vLLM engineering data (weights 451.5 to 250.1GB, KV 61,440 to 34,560 bytes per token). Price panorama: M3.1 Flash at 0.15/0.60 dollars with 0.015 cache; DeepSeek matches off-peak but charges 0.003 cache and doubles at peak; GLM-5.3 at 8/28 yuan with 2 cache; K3.1 Pro at 2/10 dollars (media-cited screenshot). The Kimi trap: K3's 12.6 on TB4.0 versus K3.1 Pro's self-reported 65.7 shipped early October 24 differ in version and harness, and K3.1 Pro weights only land before November 4. Contradictions listed without adjudication: Automation-Bench 62.2 versus 62.8, 428B versus 220B HF metadata, checkpoint sizes 250.1/233/220GB.

Oct 24, 20268 min read
Hardcore Reviews

Open Weight, But How Open? Five Model Licenses Compared

Five open-weight models crowded into the same mid-October 2026 window, all saying 'open source' - and the substance spans three tiers. Tier A, genuinely open: Reflection Beam under Apache 2.0 with zero riders and ungated downloads; DeepSeek V4.1 Flash 1017 under MIT with zero riders; NVIDIA Nemotron 3.5 Bolt 74B-A9B under OpenMDW-1.1 and the only one of the five to publish weights, training data, and recipes together. Tier B, open weights with riders: Meta Muse Spark Open carries an attribution duty that extends to distilled products (models trained on its outputs must also display Built with Muse Spark Open), a 20-million-dollar annual-revenue threshold for MaaS resale, and gated HF downloads - legally not open source. Tier C, promise unfulfilled: Mistral Large 4 had no weights as of October 23, a license with no name, and an AA classification of proprietary. On the shared ruler of AA Intelligence Index v4.3.2: GLM-5.3 at 45, Kimi K3 at 44, DeepSeek 39.5 about even with Muse at 39, Mistral 38, and Beam 30 (Beam has no public API; its score comes from self-hosted inference). Self-reported DeepSWE and Terminal-Bench numbers use different harnesses and versions and cannot be compared directly. Hardware accessibility spans two orders of magnitude, from Bolt's single 24GB card to Muse's 700GB+ NVFP4 and Beam's 294GB checkpoint (about 400GB of memory to start). The piece includes an eight-dimension license table, a watering-down spectrum (attribution > revenue threshold > gated download > late weights > unnamed license > closed data), and per-use-case protocol advice; single-model deep dives for Muse and Beam live in their own articles.

Oct 23, 20268 min read
Hardcore Reviews

FLUX 3 Max: Four Official Speed Numbers, Who's Inflating 22x?

fal Research and Black Forest Labs launched FLUX 3 Max on October 19, 2026. It is not a new foundation model but a post-trained, reinforcement-learned and inference-optimized version of BFL's FLUX 3, with fal handling post-training and systems work while BFL supplies the base. The official speed and Elo table (Elo from the preliminary Artificial Analysis AA-Video-I2V v1.1 board; generation time averaged over a 10-second 16:9 video at 720p/768p) reads: FLUX 3 Max at 1,139 Elo and 12.45 seconds, Gemini Omni Flash 1.1 at 1,134 and 47.37 seconds, MiniMax H3 at 768p with 1,141 and 83.26 seconds, FLUX 3 at 1,089 and 126.44 seconds, Wan 3.0 at 1,127 and 180.16 seconds, and Seedance 2.5 at 1,139 and 277.80 seconds, across 4,868 votes - with MiniMax H3 listed first on the board and fal's own first-place claim resting on overlapping confidence intervals. The real story is that fal publishes four different speed numbers: 12.45 seconds on its landing page, 10.95 in the news release, about 11 on the API model page, and 12 on BFL's site; the multipliers come in both 22x and 25x, and the speedup over the original FLUX 3 is quoted as both 10x and 11.5x, while fal's own pricing table lists 0.085 dollars per second next to the model page's 0.0425. More important, the 277.80 seconds for Seedance 2.5 comes purely from fal - ByteDance has never published a generation-time benchmark for that model, and fal does not say which resolution tier produced it, so the same-quality premise is not closed; independent testing found only about 10x. Mechanically FLUX 3 Max exposes a reasoning_level parameter that includes prompt expansion while Seedance 2.5 has no counterpart, fal's timing excludes prompt expansion, decoding and upload, and the two differ on reference-material limits, keyframe mechanics, and editing or continuation endpoints. Pricing: 0.0425 dollars per second promotional and 0.17 list at 720p, 0.0725 and 0.29 at 1080p, with 75 percent off running from October 15 to November 7.

Oct 22, 202610 min read