Roleplay clips are dialogue, and dialogue needs a transcript
Every other kind of gaming clip has an event at its centre. A roleplay clip has an exchange. A negotiation that turns, a character breaking in a way that is funnier in context, a threat delivered deadpan, a stranger walking into the wrong scene. None of that is loud and none of it is on the HUD. The only place it exists is in what was said.
That is why the whole stream is transcribed before anything is chosen. We read the transcript from start to finish and pick the stretches that have a beginning, a turn and an end, which in roleplay means the scene, with the line that set it up still attached.
Captions do the heavy lifting
A roleplay Short lives or dies on whether a stranger can follow the exchange with the sound off in a feed. The captions are burned in word by word, so every line lands, and the hook sits at the top of the frame to tell a scroller whose scene this is and why it is about to go wrong.
The transcript covers everyone on the recording's voice track, so an exchange between characters is captioned as an exchange, whoever was speaking.
Your facecam over the city
Roleplay streams almost always carry a facecam, because your reactions out of character are half the show. We find it, cut it out of its corner and enlarge it to fill the top of the vertical frame. The game sits underneath at full width, scaled rather than cropped, with the city, the car and the other character all still in shot. The two meet on a single seam with no blurred filler.
Getting a stream in
We fetch from YouTube, not from Twitch. If you export your VODs to YouTube, sign in with YouTube and we find them, or paste the link to one. If you do not, download the VOD from your Twitch dashboard and upload the file here. A pasted link can be up to four hours; on the free tier each of your first two videos can be up to two hours.
- Already on YouTube: sign in, or paste the video link.
- Not on YouTube: export the VOD from Twitch and upload the file.
- Long streams: cut the VOD into parts and run each as its own batch.
Post the scene from here
Give a finished clip a title and it uploads to your own YouTube channel as a Short, private, unlisted or public. Nothing is posted unless you press the button.
Last night's stream has a scene in it
Point us at the VOD and the exchanges that stood on their own are cut, captioned and waiting with your facecam in frame. The first two videos are free.
Frequently asked questions
›My streams are six hours. Can it take that?
A pasted link can be up to four hours, and on the free tier a video can be up to two. For a longer stream, split the VOD into parts and run each part as its own batch; every part is read in full.
›Does it understand that I am in character?
It reads the words and the shape of the conversation, not the lore. It does not need to know who your character is to notice that an exchange built, turned and landed, which is what it selects on.
›Will the other characters be captioned?
Yes, if their voices are in your recording. The captions follow the whole voice track, whoever is speaking.
›Does it work on a stream without a facecam?
Yes. With no facecam the clip is reframed to 9:16 the ordinary way, with the action centred. The dialogue selection and captions are the same.
›Does it find shootouts and chases?
Only when someone talked over them, which in roleplay is nearly always. It is choosing from the dialogue, so a silent chase is invisible to it and a narrated one is not.