Cast
Videos and podcasts made from your own files, contacts and chats.
Cast makes a short video, or a podcast, from what your workspace already holds. You drop in your material, say what it is for, check the script scene by scene, and Cast builds the file with a voice, pictures and captions and saves it to your Drive.
Cast makes four things:
- Video: one voice over pictures, with captions.
- Podcast video: two hosts in conversation, drawn as a video.
- Podcast audio: the same conversation as sound alone.
- Testimonial: your customers on camera, cut like a film's interview (see A testimonial).
Only the owner of a workspace makes videos and podcasts in it.
Start a new one
New opens a page laid out like the website create: a box to say what the video or podcast is about, and a wall of templates under it.
- Say what it is about in the box, in your own words. Drop what it should be made from onto the box (a Drive file or folder, a contact, an AI chat, a knowledge base or a pilot), or use Add items. Files put in with the paperclip are saved to your Drive and used the same way.
- Pick a template from the wall. Each one shows its first frame, and plays a short clip while the pointer rests on it. Drag it onto the box, or press it and then Use this template. A template sets what is made, what it is for, the look, the shape and the length; you can change each of them in the next step.
- Answer the questions, only the ones left. When you send, the questions come one card at a time: what to make, what it is for (a video), the length and shape, the look, and who speaks in which language. Each opens on what the template set, or on your settings. Cast asks only what nothing has answered yet. Words that say it (a podcast, a testimonial, 30 seconds, a minute, vertical or 16:9, a reel, a tutorial, a promo) answer that card, and words that describe the piece in a sentence or more, a template, or a YouTube video answer them all, so you go straight to the last card. Every fact on it (what it is, how long, its shape, its look, the language and the voice) opens its own card to change it.
- Or go step by step. The wand beside Your drafts, Step by step, asks every question as a wizard, starting from what to make, where each kind is shown as a frame of a piece Cast made. A pick moves on by itself. After what to make (and what a video is for) comes what it is about and what it is made from, then the other questions. Words already in the box carry over.
- Make it. The last card says what will be made, from what, and about how many credits it takes. Make it writes the script and then builds it, every scene confirmed. Show me the script first stops at the storyboard so you can check every scene before you build. Either way the studio opens on it while it works, and you can close it: a build asked for with Make it starts as soon as the script is written while the Cast page is open.
In the style of a YouTube video
Paste a YouTube link in the box, with or without words of your own, to make the new one in that video's style. When you send, Cast watches the first minutes of the video and reads how it is made: where its shots cut and how (hard cuts, fades, slides), its tempo (how long a shot runs, how many cuts a minute) and how fast it speaks, its colours, its type, its captions, how the camera moves, and how it opens, is built and sounds. The first card shows what was found, and the next is the last card, with what to make, what for, how long, its shape and its look taken from the video (its look as From the video); press any of them there to change it. The script is written to its tempo and the video is cut with its transitions.
Only its style is used: nothing of what it says, shows or sounds like is copied, and it is not kept. Reading it takes about a minute and uses a few credits. A private, members only or age restricted video cannot be read; Go on without it goes on to the questions. When you give only the link, the last card asks what the video is about.
Nothing is saved until one of the two is pressed, so closing the page leaves no draft. Your drafts lists the ones you started before.
What it is made from
A video or a podcast is made from up to 20 items. Add them in the brief, with the Add menu, or drop them on the Cast page to start a new one with that item as its first source.
- Drive files and folders. Text and documents are read: PDF, Word, PowerPoint, Excel, CSV, text, Markdown, HTML and JSON. Pictures in the files can appear in the scenes. A folder is read file by file, up to a limit.
- Contacts and contact groups. A contact gives its public card only: its name, company, role, country, tags and the public facts found about the person and the business. Never its notes, its calls or its mail. A group gives its name and the names of its members, nothing else about them.
- AI chats. What was said in the chat, and the pictures made in it.
- Knowledge bases. Cast asks the knowledge base for its key facts and figures.
- WebPilot pilots. The pages a pilot may visit, for live screenshots. It needs WebPilot to be switched on. A pilot can also be picked in the Live screenshots option, which switches the option on.
- Links. A full web address that starts with https, up to 500 characters.
- Notes. Facts, quotes or anything it should say, up to 4000 characters.
- Connectors. One of your connected accounts in the workspace, picked from the Add menu's Connector or dragged from Plugins > Connectors. A connector is kept as a source, but nothing can read one yet: it is marked as kept and unreadable, and the script is written from the rest.
The brief
The brief says what to make. Open it from the studio's first step, Brief.
- What to make: Video, Podcast video, Podcast audio or Testimonial.
- What it is for: Tutorial, Promo, Explainer, Product demo, Announcement, Social teaser, Testimonial, Onboarding, Recap, Personal message, Infographic or FAQ. Each tells the writer how to open, how to pace and how to close.
- Describe it: what it should cover, in your own words, up to 2000 characters. Together with a source this is enough to write a script, and a few words are enough on their own. With nothing described and nothing added, Cast asks for one of them first.
- Made from: what it should draw on, up to 20 items (see What it is made from).
- Language: what it is written and spoken in. Auto follows the language you describe it in.
- Voice: a voice that reads the script, or your own recording: an audio file from your Drive, such as an MP3, M4A or WAV. The script is then written from its words and keeps them.
- Hosts: a podcast has two voices. Host 1 opens and closes, and the second host is the other voice. Each voice goes by a first name of the piece's language (Hebrew names for a podcast in Hebrew, French ones in French), and the hosts call each other by those names. While the language is Auto, the names are those of the language you read the dashboard in when you made the piece.
- Speakers: a testimonial has one or two speakers instead of a voice and a main character. Each is a Real customer or a Sample actor, with a face, a name and a role on screen, and a voice of their own. Give a real customer their own name: a face from the gallery carries a name of its own. Add a second speaker for two people interviewed apart, telling the same story.
- Main character: who is seen on screen. No character, a face from the gallery, one of your Assistants, or your own picture (a JPG, PNG or WebP of one face, or your profile picture in one press). Choosing a face can change the voice to match.
- Length: a video runs 15, 30, 45, 60, 90, 120, 180 or 300 seconds, and a podcast 1, 2, 3, 5, 10, 15 or 20 minutes. Your plan caps the longest. Longer ones that your plan does not hold show which plan opens them.
- Size: Vertical (9:16, for Reels, TikTok and Shorts), Square (1:1, for feed posts), Portrait (4:5, for the Instagram feed) or Wide (16:9, for YouTube and websites).
- Look: Your brand, Dynamic, Minimal, Bold, Cinematic, Playful, Corporate, Neon, Elegant or Film (a testimonial starts on Film), each drawn as this piece's first frame in its own shape. A look sets the colors, the type and how the captions are drawn. Your brand uses the brand color saved in Branding; with none saved it cannot be picked.
- Type: which pictures the scenes use. Dynamic picks the best one for each scene, Infographic uses cards with figures, Slideshow a drawn picture or a stock clip, Screen tour live screenshots of your pages, Presenter your main character on screen, and Text words on cards.
- Options: Captions show the spoken words on the screen as they are said, in the look's own caption style or one you pick: word by word, rounded, boxed or outline. Live screenshots have WebPilot open your pages and take real pictures of them, and are off while WebPilot is off in the workspace. Stock footage uses free clips and photos where they fit a scene. Music puts a track under the voice, lowered while it speaks, from a Drive file or, where it is switched on, made for this one by AI with no singing. Say how it should sound, or leave it empty and Cast writes it from the look and what the piece is for.
A preview on the side shows the first frame in the look and shape you chose, the number of scenes or turns, the voice, and about how many AI credits the script, the pictures and the voice take, with the music where it is made by AI. The real cost is counted as it is made.
Press Create script when the brief is ready.
The scene editor
The script is written into scenes: scenes for a video, turns for a podcast (see A podcast's script). A video opens in the scene editor, where every scene is a card you watch, change and render on its own before anything is built. Scripts and pictures are AI work and use AI credits. You can close the window while it writes, and it keeps going.
The editor has three parts:
- The viewer plays the scene you picked. A scene not rendered yet shows its first picture with a likeness of its settings; Render scene makes the real one, at a small size, so you see exactly what the build will make. Play all plays the scenes one after another.
- The inspector holds the scene's settings in seven tabs.
- The strip of scenes: every scene as a card that plays when the pointer rests on it, its length, the settings it carries, and lines for its voice, music and sound effects. Between two cards, the cut opens the transition from one to the next.
What a scene can change
- Scene: its length (as long as its words, or held longer), its speed from 0.5x to 2x (the voice keeps its pitch), a speed ramp over footage (slow then fast, fast then slow, a slow-mo moment, speed up, or a freeze at the end), and what is said.
- Picture: what it shows: stock footage, a drawn picture, an illustration, a vector, a screen, a card, or the presenter as a talking head, with the HeyGen model it is made with. Picture in picture: the presenter in a corner, or a picture or footage from your Drive, round or boxed, small to large.
- Look: a filter from 23 (Vivid, Warm, Cinematic, Golden hour, Noir, Polaroid and more), Apply to all to give every scene the same one, an effect from 20 like VHS, Glitch, Old film, Light leak, Duotone or Snow, with its strength, the Ken Burns move inside a picture (10), a camera move over the whole scene (12, from Push in to Handheld and Dolly zoom), and the background behind a card or a picture that does not fill the frame.
- Text: an animated title in one of 24 styles (Rise, Typewriter, Kicker, News strap, Count up and more), and captions on or off in this scene, in their shape.
- Sound: the voice, how the line is delivered (Calm, Warm, Bright, Energetic, Serious, Soft, Storyteller or Newsreader), a silent scene, music (the piece's own, one of 32 moods made for the piece, a track from your Drive, or off here, with its level), and sound effects from 20, each at its moment. A sound effect plays as you add it.
- Extras: widgets from 18, each from its moment: Subscribe, Like, Share, emoji, an arrow, a circle, a highlight, a progress bar, your logo, a name plate, a comment, a rating, a countdown, a place, a sale badge, confetti, hearts and a spotlight.
- Cuts: the transition in and out of the scene, from 30 (Fade, Slide, Push, Whip pan, Glitch, Iris, Barn doors and more), its length, and a whoosh on it.
A scene you change shows Changed until you render it again. Render changed renders every changed scene in turn. Rendering costs only what the build would pay for the scene anyway (a voice not yet recorded, a picture not yet drawn, a talking head), and keeps it, so the build does not pay again. Everything drawn (filters, effects, cuts, titles, widgets, sound effects, motion) is free.
The words and the scenes themselves are yours to change too:
- What is said, in the Scene tab: changing it records the voice again when you render or build.
- On screen, in the Text tab: a few words shown on the picture.
- A card's own words, for a card or the opening title: its layout (a headline, a number, points or a quote) and the words of that layout. A change draws the card again.
- Who says it, in a testimonial of two speakers.
- Just a new picture, in the Picture tab: the words stay and the picture is drawn again.
- The card's menu on each scene: Move earlier or later, Rewrite this scene from a line of yours (left empty, the writer decides), Add a scene after (at your plan's limit of scenes it is off), and Delete scene (it asks first; a video keeps at least one scene).
- More, at the top: Rewrite the script writes the whole script again from a few words about how it should change, and Delete all starts over from the brief, after you confirm. This cannot be undone.
Before you build, the estimate at the top says about how many credits the build still takes, line by line: voices not yet recorded, talking heads, pictures and music. What this server already holds is not counted again.
The strip's head says how many scenes there are and how long they run together; a piece longer than the length you asked for is shortened by holding a scene less, speeding one up or deleting one.
A podcast's script
A podcast's script is turns, each said by one of the two hosts. A turn holds what the host says, which host says it (press the host to give the line to the other one), Confirm, Delete and Move. You can add a turn at any gap or at the end, up to 140 turns. A podcast turn always shows the two hosts.
A testimonial
A testimonial is cut like a documentary interview. Most scenes are an Interview: the speaker fills the frame under a film's black bars, with their name and role on a lower third. Between their answers, pictures of their story (a drawn picture or stock footage) run while they keep talking, and a card can hold a pull quote or a figure they gave. Every scene says who speaks it.
- A real customer says only what they told you. Describe it with their own words, or add their review, an email or your notes, and the script keeps to those. Their interview is marked Dramatized with AI on screen, because the voice is made for them.
- A sample actor is someone made up for a draft or a demo. Their story may be invented, and every interview scene carries Sample, not a real customer. A testimonial with a sample actor can be built, watched and downloaded, but it can never be published: not by link, email or a connected service.
Where talking heads are switched on for the server, each interview is the speaker talking to camera, made from their face and the scene's own voice, and the build shows the step Making the talking heads. Where they are not, the speaker's portrait moves in slowly while they speak.
Building
When every scene is confirmed, press Build, or Confirm all and build. In the scene editor, Build asks first, with the estimate, and confirms every scene for you. The build records the voice, prepares the pictures, puts the scenes together, mixes the sound and saves the file to your Drive. The Result step shows each phase and a count while it works.
- You can close the window while it builds. The build goes on and the file is there when it is done.
- Stop ends it. What was made so far is thrown away.
- One workspace holds at most two builds at a time, one building and one waiting. A third is refused until one of them is done.
- A build that did not finish says why, and offers Try again or a way back to the storyboard.
- A build that is finished and saved counts as one of the month's builds. A build that fails or is stopped does not.
- Recorded voices are kept for two days. Build again in that time, after a stop, a failure or for another copy, and a scene whose words and voice have not changed uses its recording again instead of paying for a new one.
The result and publishing
A built video or podcast shows its player, its length, size, voice, when it was built and its file. From there you can Publish, Download, Open in Drive, Edit (back to the scene editor, or a podcast's script), Duplicate (a draft copy with the same brief, sources and script) or build again. After you edit, the file stays as the last build until you build again.
The Publish window has four ways:
- Link. Create link makes an address. Anyone with the link can watch (or listen) until you delete the file. Treat it as public.
- Email. Write up to 10 addresses and an optional note of up to 1000 characters, then Send. They get the link, never the file. It is sent through your connected mailbox, so one has to be connected.
- A connected service. A service you connected in Plugins, Connectors with Read and publish, such as YouTube, is listed first. Publish to it opens a short form: the title (the piece's own to start with), a description, and who can see it (only you, to start with). Publish posts it through your own account, and that press is the approval. The upload runs where the connection runs: on Xuda's servers, or on your workspace server if you picked it when you connected the service. The window says when it is there, with a link to open it, or why it was not published. A video goes to YouTube; a podcast with sound only does not.
- Connector. Hand it to the AI chat: the chat gets the file and its link, you say where to post, and it asks you before it posts. Use it for a service that has no card of its own above.
A quick video
Drag Cast onto a contact, a contact group, an AI chat or a knowledge base, and a small menu opens at that spot:
- Quick video starts an infographic video about it, with the defaults saved in the Cast settings and a voice in your language. It tells you when it is ready, with a Watch button. If the settings say not to build right away, it stops at the storyboard for you to check the script.
- Open in the studio opens a new brief with that item as its first source, to pick the look and check the script first.
The menu says what is read about the item: a contact's name, company and public facts (never notes, calls or mail), a group's name and its members' names, what was said in a chat, or the key facts of a knowledge base.
The Cast page
With no videos or podcasts yet, the page shows one box with the way to make the first. After that, up to twelve show as cards, each with a print of its first frame, its state, its name, its length and when it changed. Past twelve they become a grid with search by title, filters on state and on what it is (a video or a podcast), sortable columns, a Columns menu, and ticks to delete many at once.
A line at the top says how many you have and how many are being made, waiting for you or failed. At its end, quietly, it says how many builds the plan has used this month. The states are Draft, Writing, Script ready, Building, Ready and Failed.
A card's menu opens it, duplicates it or deletes it. A delete asks first and takes the brief, the storyboard and the pictures with it. A file already saved to your Drive stays there. You can keep up to 200 videos and podcasts.
If Cast is switched off, your videos and podcasts are kept and nothing new is made until you switch it on. If the server cannot build videos yet, you can still write and edit scripts and build later.
The settings window
Settings, at the end of the page's header, holds the defaults that every new video, podcast and quick video starts from: language, voice, second host, main character, length, size, look, captions, music, the Drive folder where finished files are saved (a folder of its own, not the top of the Drive, by default /Cast) and Build a quick video right away. Nothing is changed until you press Save.
Limits and cost
- A plan holds a number of builds a month, a longest video, a longest podcast, a number of scenes in a video and a resolution. See plans and billing.
- At the month's limit you can still write and edit scripts. Only a build is refused, and the count starts again next month.
- Scripts, pictures and voices use AI credits on every plan, as in a chat. When credits run out, the work stops and says so. Add credits to carry on.
- Scenes with live screenshots need WebPilot, and AI music needs to be switched on for the server.
- A testimonial's talking heads, and a video's presenter, need to be switched on for the server, and use AI credits by the second of video they make. The estimate counts them where they are on.
- A scene rendered alone is kept on the server for two days. Music moods are made once for the piece and kept for two days too; a scene rendered alone plays a mood only once it has been made.