Dub Together gameplay turns a video clip into a new voice performance made by your group. In Parallel, everyone records a separate version, watches the results and votes for a favourite. In Character, players take individual roles and create one shared scene without a vote. The official description supports online groups of up to six, and finished dubs can be saved as videos.
Choose the result you want
Begin with the kind of performance your friends want to make, because the two modes organise the same activity differently. Parallel gives each player the complete clip and keeps the recordings separate until everyone watches them. Character divides the dialogue into roles and brings the performers together inside one result. Character requires at least two players, so include that condition when organising the group.
A useful way to choose is to ask whether everyone wants to reinterpret the whole scene or contribute one part of a conversation. Parallel suits the first idea because every participant receives the same starting clip. Character suits the second because a performer is assigned to each character. These are suggestions for arranging a session; the game's documented difference is separate versions with voting versus one version without voting.
Record your own Parallel version
Parallel starts by playing the clip without its original voices, while the text for your lines appears on screen. Speak the text into your own microphone and fit the delivery to what is happening in the picture. You record line by line without hearing the interpretations the other players are preparing. The reveal therefore happens during playback, rather than during the recording stage.
Treat the picture and displayed line as your immediate cues, and decide how you want the voice to relate to the character's actions. A calm delivery and an exaggerated delivery can suggest different interpretations of the same footage. You can choose your performance freely, but there is no documented scoring formula for acting accuracy. The official loop describes a group preference vote after everyone has had a chance to watch.
Watch every version and vote
Once the recordings are finished, Parallel shows each player's version over the original picture in turn. Listen to the whole performance before choosing a favourite so that your choice reflects the version you actually heard. Every player then picks a favourite, and the best version appears on the podium. The description does not establish a separate accuracy score, prize catalogue or reward code for winning.
The September 15 announcement gives the host control over watching a version again and advancing to the next version. That makes the host's playback decisions part of organising the group's reveal. Agreeing to finish the current version before moving on is a practical way to keep the session understandable. If you want the audience involved, the documented Twitch integration adds viewer votes through !vote or a number.
Build one scene in Character mode
Character begins with role assignments, with one performer for each character in the clip. During recording, the group follows the dialogue together and the player whose line is due speaks while the others watch. A role that nobody takes keeps its original voice. This lets you understand an unfinished assignment without assuming that every audible character must be a member of your lobby.
At the end, everyone watches the shared scene in one piece instead of comparing separate versions. There is no voting phase because the group has made only one result. Plan your delivery around the other roles' replies if you want the scene to feel like a conversation. Workshop role counts help describe a pack's dialogue, but a tag of 6+ roles does not expand the six-player lobby limit.
Use the recording window
Both modes allow you to repeat a recording that you are unhappy with while the round timer is still running. The official description does not give a universal number of seconds for this window. Check the timer displayed in the actual round before committing to another attempt. Claims about a fixed countdown, extra attempts after it ends or an unlimited practice phase remain 待确认.
Use a repeat for a specific change that you can hear or describe, such as a missed word or a different character voice. That gives the next attempt a clear purpose instead of requiring a perfect performance from the first take. Leave your choice grounded in the time the game displays, because the documented permission to redo is tied to that running timer. You can compare the result later when the group watches the completed recording.
Shape the voice and timing
The official listing documents six voice effects, with Chipmunk, Monster, Echo, Robot and Phone named in the description. You can select an effect for each line, and it is applied to the recording that the group hears during playback. The sixth effect's name is 待确认, so do not build a performance plan around an invented option. A noise filter also removes background noise between sentences, according to the developer's description.
The September 8 update introduced timing compensation with a default VOICE TIMING value of 150 ms. A higher value starts the recorded voice earlier, while a lower value adds delay. The September 15 update made the timing adjustment clearer and improved microphone quality and protection against distorted recordings. Treat 150 ms as the announced default; the best setting for your own audio arrangement is not established by these notes.
Choose a clip for your group
A pack provides the video and lines that the group uses for a round, and the game includes a Pack Creator for preparing your own material. The only bundled pack named in the official listing is an excerpt from Sprite Fright by Blender Studio under CC BY 4.0. Other packs are created by players and hosted through Steam Workshop. A familiar film appearing in a community video does not establish that it is bundled with the game.
Workshop filters documented on September 6 cover language, number of lines and number of roles. Use those filters to look for dialogue your group can read and a role arrangement you want to perform. Subscribed packs appear without restarting, and a player missing the selected pack gets it downloaded when the round starts. For editing the footage itself, continue to the clip creation guide, which explains the documented creator workflow.
Save the performance and continue
The official description lets you save finished versions as videos, including the shared result from Character. The September 15 announcement also says dubs are automatically saved to Videos\Dub Together. Use that published location as the first place to look for a saved performance on the documented Windows setup. The sources do not establish a file extension, export resolution or a guaranteed outcome for every device, so those details remain 待确认.
For another round, decide whether you want to keep the mode, change the roles or pick a different pack. That decision follows the actual distinction between individual interpretations and a shared conversation. Keep the lobby within the supported six players and let the host coordinate the next playback reveal. The online guide covers Steam joining, lobby codes and Discord invitations when you are ready to organise the group.
What players ask
What do you actually do in Dub Together?
You speak new lines over a video clip with your own microphone. Parallel produces separate versions that the group watches and votes on. Character assigns roles and combines the players into one shared scene. Both modes let you save the finished dub as a video.
Can I redo a bad line?
The official description allows recordings to be repeated while the round timer is running. It does not publish a fixed duration for that recording window. Use the timer visible in your round when deciding whether to try again. The available time in seconds remains 待确认.
Does every mode end with a vote?
Parallel has a vote after the separate recordings have been shown. Its winning version appears on the podium. Character creates a single shared result and has no vote. Choose the mode according to whether your group wants separate interpretations or one dialogue.