02 Creating a Project
DeepVideo has four project types, and they differ more than their names suggest. Picking the wrong one sends you on a long detour, so read this section first.
1. The Four Project Types
| Type | What it does | Needs internet? | Extra switch required? |
|---|---|---|---|
| Expert Video Translation | The full chain: transcribe the original speech → translate it → dub it in the original voice → burn in subtitles → assemble the finished video | Only for the translation step | No |
| Quick Video Translation the entry in the left sidebar reads "One-Click Video Translation" | Also delivers a finished video, but exposes fewer options and caps the quality tiers | Only for the translation step | No |
| Video Dub | Keeps the picture and replaces what was said with a dub in the same language that says something else; you can import your own subtitles | Yes | Turn it on under Settings → General → Experimental Features |
| Video Subtitle | Produces subtitles and burns them into the picture, no dubbing | Yes | As above, the experimental switch is required |
To have a video speak a different language, choose Expert. To get subtitles only, or to burn subtitles only, choose Video Subtitle. If you already have a translated video and want to rewrite the lines and re-record them, choose Video Dub.
2. Options In The New-Project Form
The list below follows the groups as the interface groups them, and every item states its default value. The default is the value you should leave alone the first time you use the app.
File & Language
- Project name: call it whatever you like, you can rename it later.
- Video File Path: accepts
mp4 / mkv / avi / mov / webm / flv. It can also be a platform video you already have on disk — a file you fetched with Video Download. - Source Language (default Auto-detect): the language spoken in the video. Auto-detect is good enough most of the time; if you know what the video speaks, select it yourself and be certain.
- Target Language (default English): the language you translate into. Video Subtitle and Video Dub have no such field.
Quality Tiers
- Speech Recognition Quality: pick from Fast / Standard / Professional. Higher is more accurate and slower — the app's own hint says "Higher quality models are slower but more accurate". For a long video, running the Fast tier once to see what you get costs far less time than sitting through a Professional run to the end.
- Speech Synthesis Quality: pick from Fast / Audio Quality (the Expert type adds the higher Professional tier; Quick Video Translation does not have it). Higher sounds more natural, and is slower. Mind that the second tier is labelled "Audio Quality" in the interface — "Standard" is the name used by the Speech Recognition Quality field above.
Voice
One of three, and the default is Voice Clone:
- Voice Clone: clones the timbre of the person speaking in the video and has that voice speak the target language. The option that stays closest to the original.
- Built-in Characters: the set of voices that ships with the app.
- Custom Role: roles you created yourself on the Voice Roles page.
If your plan does not cover cloning, this choice is grayed out and shows "Voice cloning not supported in your current plan"; the form then falls back to Built-in Characters by itself.
Expert Video Translation does not choose a voice here — it always runs in two stages and the voice is set in stage two, in the left sidebar of the editor, so this field is not in the form at all. One-Click Video Translation, Video Subtitle and Video Dub still ask you here.
Subtitles
The Expert Video Translation form has none of this group (it is set in the editor's left sidebar as well); the items below appear in the other three types:
- Add Subtitles (default off): decides whether the finished picture carries burned-in subtitles at all. Turn it on if you want subtitles — this is the option people miss most often.
- Subtitle Language: Target only / Source only / Bilingual.
- Subtitle Order: with Bilingual, which line sits on top — the two choices are named exactly Source on top and Translation on top.
- Subtitle Size (default Medium), Subtitle Background (default No), Font (with Bilingual this becomes two fields: Source Font and Target Font), Subtitle Position (three choices: Auto Detect / Bottom / Top; Auto Detect is the default).
- AI Subtitle Correction (default on): a large model fixes the wrong characters the speech recognizer produced. Switching it off saves a little cloud usage, at the price of carrying mistakes in the source text straight into the translation.
Multi-speaker & Expression
- Multi-speaker mode (off by default, forced on for Video Dub): detects how many people speak in the video, gives each of them a different voice, and lets you edit per person. The biggest difference it makes is in the steps — with it on the pipeline runs one extra Speaker Identification step, see 03 Translation Pipeline & Progress.
- Number of speakers: Auto-detect, or a fixed 2–5 speakers. When detection keeps getting it wrong, naming the count yourself is less work.
- Emotion Recognition (off by default): reads the emotion in the original voice and uses it to shape the delivery of the dub.
Volume
- Voice Volume Adjustment: −50% to +50%, which sets the dubbed voice against the background audio.
- Loudness Normalization (on by default): evens out the overall loudness. The Expert type does not ask in the form — you set it in the editor.
- Preserve Source Volume Balance (on by default): keeps the voice-to-background ratio of the original video as closely as it can. Only the Expert type has this, and it is set in the editor too.
Translation Control
- Enhanced Translation (off by default): switch it on and you can set Translation Style, Industry Domain and Custom Instructions.
- Use Glossary (off by default): attach one or more glossaries to lock down how brand names and proper nouns are translated.
It always runs in two stages: stage one runs up to the "Translation" step and stops there, then the editor opens by itself and waits for your review. Dubbing and assembly continue only when you click "Generate Video" in the editor. If you want one uninterrupted run, choose Quick Video Translation; the Batch Queue, which queues many videos at a time, runs that same type too (see 08 Project List & Batch Queue) and never stops for a person either.
Yes. Subtitle style, font, position, background, volume and voice are all changeable at any time in the parameter panel on the left of the editor — you do not need to rebuild the project. But Source Language, Target Language and Speech Recognition Quality, which decide what gets recognized, are read-only there: to change one of those, go back to the new-project form and start a new project.
3. What Quotas Stop You From Running
Two hard lines: a single video whose duration passes your plan's cap is refused outright (1 minute on the free plan), and you are refused once the cumulative duration in your billing period is used up. The exact figures and the message you get when you hit either line are in 10 Plans & Quotas. File size carries no extra limit of its own — only duration counts.