#7723 como remover pausas de falas?

To remove silent pauses between dubbed lines, use the built-in options in the Advanced Options panel.

1. Open the top menu: Menu -> Tools/Options -> Advanced Options.
2. Go to the Dubbing Settings section and enable Remove the mute buffer each subtitle audio (this trims leading/trailing silence in each dubbed clip, improving sync).
3. Go to the Alignment section: enable audio speed-up and optionally video slow-down so the engine compresses gaps between subtitle segments to "borrow time" and avoid overlaps.
4. Also consider enabling Secondary recognition (recogn2pass) to regenerate tighter, better-aligned subtitles.

Note: the alignment engine automatically removes/compresses inter-subtitle silent gaps when audio speed-up and/or video slow-down are enabled. If sections still overlap or drift, make sure you check Clear Output before retrying.

Please refer to the relevant documentation:

• Subtitle / audio / video alignment
• Advanced Options

mas essas opçoes ja são de padrão do programa , o texto o time não tem pausa , mas a dublagem esta deixando com pausas

The issue is that TTS engines (especially Edge-TTS) add leading and trailing silence to each generated clip, and slight over-generation makes the audio shorter than the subtitle window, leaving a gap. This is not a translation/text problem — it is a per-clip silence buffer problem.

1. Enable Remove the mute buffer each subtitle audio (menu: Menu -> Tools/Options -> Advanced Options -> Dubbing Settings). This trims the auto-generated silence at the start and end of every dubbed line, which is the main cause of the gaps you hear.
2. Enable audio speed-up (Alignment section) so when a dubbed clip is shorter/longer than its subtitle window, the engine stretches/compresses to fill the exact time slot.
3. Enable video slow-down together with audio speed-up — combined, the engine ignores per-item limits and distributes the timing difference, closing silent gaps.
4. Increase 配音语速 / Speed+% (e.g. +10 to +20) in 设置更多参数 — longer clips leave less residual silence.
5. Enable 二次识别 (SecondarySTT) — it re-transcribes the dubbed audio and regenerates short subtitles that align tightly with the actual speech, removing perceived pauses.
6. If you are using clone voice, ensure 最短语音持续 is at least 3000ms, otherwise clips get cut with extra silence.

Also note: after changing any of these, check 清理已生成 (Clear Output) before retrying, otherwise old cached dubbing audio will be reused.

请查阅相关文档：

• Subtitle / audio / video alignment
• Advanced Options

essas opçoes ja são do padrão do programa , isso não da certo , ainda esta tendo pausas

If these options are already enabled by default and you still hear pauses, the cause is not the alignment engine — it is the source of the pause itself. Please confirm the following so I can pinpoint it:Are the pauses at the END of each dubbed line (a silent tail before the next line starts), or BETWEEN words inside a single line?Are you using a local built-in TTS (Qwen3-TTS / F5-TTS / OmniVoice / Confucius) or an online API (Edge-TTS / Doubao / etc.)?Which 配音渠道 (TTS channel) and 配音角色 (voice role) are selected?Is 二次识别 (SecondarySTT) enabled or not?What is the exact software version (shown top-left of the window)?Please provide the above, plus the last ~30 lines of the latest log file in the logs/ folder (软件目录/logs/, e.g. 20250622.log).暂无法确定具体原因。请提供软件安装目录下的 logs/ 文件夹中最新的日志文件（复制底部约30行报错信息）以及上述信息以便分析。请查阅相关文档：FAQTTS output issues

não esta resolvendo, quando termina um bloco para a outra a fala esta tendo pausa

The pause you are hearing is the natural dead time between one subtitle's time window end and the next subtitle's time window start, i.e., the original SRT timeline has a gap between blocks, and the dubbed audio simply plays inside that window and stops, leaving silence until the next clip begins. The alignment engine does not remove these baseline gaps unless the subtitle timings themselves are changed.Use the CLI/Advanced parameters that control subtitle timing rather than dubbing: in Advanced Options -> ASR Settings, reduce 静音分割持续毫秒 (VAD: Min silence duration for split(ms)) so the ASR splits on shorter pauses and produces blocks with less dead space between them.Ensure 最长语音持续 (max speech duration) × 1000 is greater than 最短语音持续 (min speech duration in ms), otherwise the splitting logic behaves inconsistently and leaves gaps.If you already have an SRT, edit it externally (menu: 打开字幕文件夹 in single-video mode) and shorten the gap between blocks by setting each block's end time closer to the next block's start, then click 不保存只继续 so the software uses your edited SRT.Enable 二次识别 (SecondarySTT) — it re-transcribes the dubbed audio, so the final subtitle blocks match the spoken audio and eliminate residual silence between blocks.Note: the alignment engine can only compress or borrow time between blocks when audio speed-up / video slow-down are active AND there is an actual overlap or excess to distribute; pure baseline silence gaps in the SRT are not removed by default.This problem is best diagnosed with your actual files. Please provide: software version (top-left), TTS channel and voice role, whether 二次识别 is on, and the last ~30 lines of the newest file in 软件目录/logs/.暂无法确定具体原因。请提供软件安装目录下的 logs/ 文件夹中最新的日志文件（复制底部约30行报错信息）以便分析。请查阅相关文档：Subtitle / audio / video alignmentVAD speech recognition tuningSingle-video mode

o texto não tem pausa, o time e em seguencia sem pausa, mas o audio esta criando pausas

This narrows it down: the text has no pauses and the timeline is continuous, yet the generated audio clips contain silence — that is TTS-engine-side padding, not alignment, not VAD, not translation.

The most common causes and fixes:

1. If you use Edge-TTS: it inserts silence at the start/end of each synthesized segment. Enable 移除配音前后静音缓冲 (Remove the mute buffer each subtitle audio) in Advanced Options -> 字幕配音调整 (Dubbing). If it is already on but clips are still padded, switch to a different TTS channel to test (e.g. Qwen3-TTS or Doubao) to confirm it is Edge-TTS padding.
2. If you use a local cloned TTS (F5-TTS / OmniVoice / Index-TTS / CosyVoice): the reference audio or the model itself appends a tail of silence, especially when the reference clip has trailing silence. Trim your reference audio and ensure 移除配音前后静音缓冲 is enabled.
3. The alignment engine only removes silence when 音频加速 or 视频慢速 is active and there is excess length to distribute. If the SRT window equals the spoken length (continuous), the engine has nothing to compress and the TTS padding stays. So test by enabling both 配音加速 and 视频慢速 explicitly and re-run, with 清理已生成 checked.
4. Also check 配音语速 (+%) — a slightly higher rate leaves less room for the engine's padding to show as a gap.

To identify which is happening I need concrete data. Please provide:

• Software version (top-left of the window)
• Which TTS channel and voice role are selected
• Whether the same pause appears when switching to another TTS channel (e.g. Edge-TTS vs a local one)
• The last ~30 lines of the newest file in 软件目录/logs/

暂无法确定具体原因。请提供软件安装目录下的 logs/ 文件夹中最新的日志文件（复制底部约30行报错信息）以便分析。

请查阅相关文档：

• TTS output issues
• Subtitle / audio / video alignment

a pausa esta com faixa de 600ms de pausa

A 600ms gap is not random TTS padding — that value matches the ASR silence-split parameter exactly. The blocks are being split with ~600ms of dead space between them (the default 静音分割持续 / VAD: Min silence duration for split = 600ms), so each subtitle ends 600ms before the next begins, and the dubbed audio leaves that gap silent.Fixes:In Advanced Options -> 语音识别参数 (ASR Settings), reduce 静音分割持续毫秒 (VAD: Min silence duration for split(ms)) from 600 to a smaller value (e.g. 200-300) and re-run with 清理已生成 checked, so blocks are cut closer together with less dead space.Ensure the alignment engines are truly active: enable 配音加速 (audio speed-up) AND 视频慢速 (video slow-down) together, then re-run. The engine compresses inter-block silence only when there is excess length to distribute.If you already have an SRT with continuous timing and no gap, import it directly (put it in the _video_out/[name]-[ext]/ folder as the language code, do not check 清理已生成) so ASR is skipped and your continuous timeline is used instead of the 600ms-split output.If the pause persists after lowering the split value, enable 二次识别 (SecondarySTT) to regenerate tightly aligned blocks from the dubbed audio.If this still does not solve it, I need the log. Please provide:Software version (top-left)Whether the 600ms gap appears in the generated source SRT (open 打开字幕文件夹 and check the timestamps) or only in the final audioThe last ~30 lines of the newest file in 软件目录/logs/暂无法确定具体原因。请提供软件安装目录下的 logs/ 文件夹中最新的日志文件（复制底部约30行报错信息）以便分析。请查阅相关文档：VAD speech recognition tuningAdvanced OptionsImporting local SRT

cada pontuação esta deixando uma pausa de 600ms ou mais , sendo que deveria ser um áudio continuo sem pausas, olha o exemplo de como e meu texto e esta ficando com pausas

1
00:00:00,080 --> 00:00:20,280
Ao ouvir a declaração de amor do Mark, JJ fica bastante mexida com toda aquela situação. No entanto, ela logo percebe que não quer apressar as coisas e percebe que ainda não tem certeza sobre os seus próprios sentimentos. Para escapar daquela pressão, ela desconversa, transforma tudo em uma grande piada e foge dali.

Now the cause is clear: the pause is not between subtitle blocks — it is inside a single subtitle line. The text contains commas and periods ("...situação. No entanto...", "...e foge dali."), and the TTS engine treats each punctuation mark as a sentence/clause boundary and inserts its own silence there, stacking up to 600ms or more, so one continuous block sounds chopped.The fix is to remove the punctuation that triggers the split, or reduce text normalization:In Advanced Options, find 文本规范化 (Text normalization) in 字幕配音调整 (Dubbing). Turn it OFF and test — normalization can expand punctuation into pause-triggering tokens.Best fix: remove the intermediate punctuation from the subtitle text so the entire line is read as one flow. In single-video mode, edit the subtitle line in the editing window (double-click the 字幕 cell) and delete the internal 。,，. ！? ! ? — or replace them with a space — then right-click the line and 重新配音 (re-dub) to hear the result before continuing.Alternatively edit the SRT externally via 打开字幕文件夹, strip internal punctuation, save, then click 不保存只继续.Switch to a different TTS channel to confirm the behavior: Edge-TTS is the most prone to inserting clause pauses at punctuation. Doubao 2.0, Qwen3-TTS, or a local clone model usually respect the whole line as one utterance and produce a more continuous read.If you keep the punctuation but want it read faster/continuous, increase 配音语速 (+%) and keep 移除配音前后静音缓冲 enabled — but this only trims clip edges, not the internal comma pauses, so removing/seconding the internal punctuation is the real fix.Note: the 600ms figure is a coincidence with the VAD value — the internal pause here comes from the TTS punctuation handling, not from VAD, because the whole block is one subtitle line.请查阅相关文档：TTS output issuesEdge-TTSSingle-video mode editing

realmente a pausa vem do " . " o problema que o " . " ele deixa uma pausa de mais de 600ms , isso e muito , a pausa realmente precisa ter

Understood — the punctuation pause is legitimate, but the length (600ms+) is too long. In pyVideoTrans there is no direct setting that adjusts the internal punctuation-pause duration of the TTS engine; that pause length is controlled by the TTS engine itself, not by the software.What you can actually do:Switch the TTS channel. Pause length at punctuation is engine-specific. Edge-TTS produces notably long clause pauses; Qwen3-TTS (内置), Doubao 2.0, Confucius-TTS, OmniVoice, or a local clone model (F5-TTS / Index-TTS / GPT-SoVITS / CosyVoice) usually produce shorter, more natural punctuation pauses. Test the same line on another channel — this is the only reliable way to change the internal pause length.Use the single-video editing window to re-dub the line and preview: open the block, and if the pause is still too long, right-click the line and 重新配音. If a specific engine cannot be tuned, changing engine is the answer.Keep punctuation but shorten the perceived pause by combining: enable 移除配音前后静音缓冲 (removes clip-edge silence) AND raise 配音语速 (+%) slightly — this shortens overall read time but does not shorten the punctuation pause itself, so the effect is limited.Practical workaround for the internal pause: replace " . " with a shorter pause trigger acceptable to the engine, e.g. "，" (comma) or a line break within the SRT — many engines pause less on a comma than on a period. Try replacing periods with commas in that line and re-dub to compare.If you want the pause preserved but shorter, the honest answer based on this software: there is no built-in parameter to set the exact millisecond length of a TTS engine's punctuation pause. It is determined by the TTS channel. Switching channel is the solution.请查阅相关文档：TTS output issuesAll TTS channelsSingle-video mode editing