LOCKHEART STUDIO • TEST LAB

Your first video,
one step at a time.

A plain-language guide for phone and desktop testers.
Open → Plan → Create images → Animate → Review → Export

Before you start

This guide is for testers using the hosted Studio from a phone, tablet, or computer. The host computer does the image generation, animation, lip-sync, and export work. Your browser is the remote control: you do not need to install AI models or run Windows BAT files.

Have an audio file saved on your device, a short idea for your video, and optionally a clear character reference you have permission to use. For your first project, use a short audio excerpt and a few scenes. A full song can require many individual renders.

Read Discord’s #rules and #announcements first. Get the current access code from #latest-build. This guide deliberately leaves the code out so it can be replaced without changing the manual.

Back to contents ↑

1. Open the Studio and sign in

Open your browser. Chrome is a straightforward choice on Windows or Android.
Open https://studio.lockheartstudio.com.
Enter the current code from Discord’s #latest-build, then choose Open the studio.
Wait several seconds for the dashboard to load. Do not tap the login button repeatedly.
Read the active project name before changing anything. A shared Studio may display the last tester’s project.

If it says “The studio is currently in use”: another remote tester has control. Wait for their turn to finish. Do not try another browser to bypass it. The update allows takeover after five minutes of inactivity; availability does not mean another person’s project belongs to you.

If the website will not open: check #announcements. The permanent address still depends on the host being online. Report the error instead of installing anything.

Back to contents ↑

2. Understand what you are looking at

Project: The workspace holding a song, scene plan, prompts, images, and animation takes.
Scene: One short shot in the finished video. Several scenes are assembled into the complete song timeline.
Prompt: A written description telling the Studio what to create.
Starting image: The still picture used as the visual starting point for an animation.
Take: One attempt at a scene’s animation. A new take is another attempt.
Render / generate: The process of creating media. This takes time on the host computer.
Queue: The waiting list of jobs. Queued means waiting, not broken.
Master / export: The assembled video file you download at the end.

The top bar identifies the current project. System cards show whether image and video tools are ready. Scene cards hold each scene’s prompts, previews, status, and controls. Batch controls work on multiple scenes. The export panel builds the finished video. The event log records activity and errors.

Statuses: Ready = available for editing; Queued = waiting; Generating/Rendering = working; Completed = a take is available; Approved = accepted; Rejected = needs correction; Failed = read the error before trying again.

Back to contents ↑

3. Create your first project

Choose New project. Do not overwrite someone else’s project.
Enter a clear project title, such as Alex — First Test, and the artist name.
Choose your song file from your device. The documented upload formats are MP3, WAV, FLAC, M4A, AAC, and OGG. A streaming-service link is not the same as an audio file.
Wait until the upload finishes. Keep the page open and do not refresh during the upload.
Set the average scene length. Four seconds is the documented default. Shorter scenes mean more shots and more rendering work.
Set title and outro lengths. The documented defaults are five and three seconds. Use zero for a section you do not want.
Describe the overall visual style, then fill in the character identity lock using the examples below.
Create the project and wait for the scene list to appear. Confirm the title, song runtime, and scene count look reasonable.

Visual style describes the whole video’s look:

Cinematic anime, midnight blue and neon green lighting, rainy city streets, moody atmosphere, detailed environments, gentle camera movement.

Character identity lock describes the features that should remain the same:

Adult performer with short black hair, neon green eyes, a black hooded jacket, and a fitted black mask. Keep the same face proportions, hair, outfit, and mask design throughout.

Keep actions such as running or turning out of the identity description. Those belong in individual scenes. Changing a scene duration later can rebalance the whole song timeline; review timing again after doing so.

Back to contents ↑

4. Plan the video with Story Director

Open Project setup.
Paste the complete lyrics, keeping verse and chorus labels when available.
Write the visual story concept: who is in the video, where they are, what changes, and how it ends.
Choose the image look and intensity. The documented intensity scale is 1–10; higher values request a more dramatic plan. It is not an animation-quality setting.
Leave Rewrite scenes unchecked if you want to protect prompts already written. Check it only when intentionally replacing the scene plan.
Choose Direct every scene and wait for the scene prompts to fill in.
Read through the scenes. Correct anything that does not match your story before generating media.

A lone performer walks through an empty rainy city. The early scenes feel isolated. As the chorus begins, green lights appear in the buildings. The ending shows the performer reaching a rooftop at sunrise.

This is a written shot plan, not the finished animation. You will create and review images and motion afterward.

Back to contents ↑

5. Add a character reference

In Project setup, find the reference-image area.
Upload a clear picture of the intended character that follows the server rules.
Check which tile is marked PRIMARY. Tap another tile if that should be the main reference.
Use a front or three-quarter view with a clearly visible face and recognizable clothing. Avoid tiny faces, heavy distortion, text, and conflicting outfits.

The primary reference helps future generated starting images stay consistent. It does not automatically replace pictures already created. To update an existing scene, request a New starting image after reviewing its prompt.

If identity keeps drifting, report the project, scene, and example to the host. Reference-strength controls are host settings; remote testers should not install or change the engine.

Back to contents ↑

6. Edit one scene before generating

Find the first scene you want to test and choose Edit scene.
Give it a short scene title that helps you recognize it.
Review its timeline duration. This is the scene’s allotted share of the song.
Fill in the starting-image prompt: describe one still picture.
Fill in the animation prompt or movement field: describe what moves during the shot.
Enable This is a visible singing or performance scene only if the character will visibly sing.
Choose a transition direction if needed, then save the scene before generating.

Starting-image example:

Medium shot of the adult hooded performer standing on a rainy rooftop at night. Neon green eyes and fitted black mask, city lights behind, hands resting at their sides, centered composition, room above the head.

Animation example:

The performer slowly turns toward the camera. Gentle breathing and slight jacket movement in the wind. Slow camera push-in. Keep the face and clothing stable.

Use one main action and one gentle camera movement. Avoid asking a short scene to run, spin, fight, transform, and zoom all at once. If the still picture is wrong, fix the image prompt; if the picture is right but the movement is wrong, fix the animation prompt.

Back to contents ↑

7. Make and check a starting image

Choose Generate image on your test scene.
Wait while the scene is queued and then generated. One click is enough.
When the picture appears, inspect the face, hands, limbs, clothing, background, and framing.
If it is wrong, edit the prompt to describe the correction and choose New starting image.
Continue only when the picture is suitable for the motion you want.

Example correction: “Keep both hands visible at the performer’s sides; exactly two arms; no extra fingers.” Specific corrections communicate the problem better than “make it better,” although a prompt cannot guarantee a perfect result.

Upload image lets you provide a PNG, JPEG, or WebP starting picture. It replaces the image used for future animation attempts on that scene. Follow the server’s Lockheart Studio-only sharing rule; use permitted original/reference material, not other generators’ output as a showcase.

For later: check scene selection boxes and choose Generate selected to work on specific scenes. Generate missing queues scenes without pictures that have usable prompts. These can create many jobs—learn with one scene first.

Back to contents ↑

8. Animate your test scene

Confirm the scene has a starting image and a saved movement prompt.
Choose Animate.
Watch the status. Queued means another job is ahead of it; Rendering means the host is working.
When completed, play the entire clip. On a phone, use full-screen playback if the preview is small.
Check that the character stays recognizable, the movement is intentional, and the background does not melt or warp.
If needed, adjust the movement description and choose New animation take.

For lifeless motion, add one clear action such as a slow head turn. For chaotic motion, remove actions and simplify the camera direction. A malformed starting image usually needs replacing before another animation attempt.

There is no reliable fixed render time in this guide. Resolution, scene length, queue size, and host settings affect the wait. Do not interrupt solely because a job is slow.

Back to contents ↑

9. Approve or reject a take

Watch the complete clip, including the beginning and ending.
Choose Approve if you accept the current result.
Choose Reject if it needs correction. Enter specific feedback when the interface offers a feedback field; otherwise record it in #general-feedback with the project and scene number.
Correct the image or movement prompt as appropriate, save, and request a replacement.
Watch the replacement. Redo lip-sync if you replace an animation that previously had lip-sync.

Scene 3: The face changes after two seconds. Keep the reference face and black jacket unchanged. Reduce the head turn and keep the camera almost still.

Reject is a review decision, not a guarantee that the next attempt will be correct. Avoid repeatedly generating the same unclear request. If the same defect continues after specific corrections, send examples to the host.

Back to contents ↑

10. Expand to more scenes and use the queue

Once one scene looks good, test a few different shots. Run test scenes selects up to five spaced scenes that already have images. It is useful after a small first test, but can queue several jobs at once.

Animate selected: animate the checked scenes that have images.
Animate unfinished: queue image-ready scenes not completed or approved. This can be a large batch.
Pause: let the active job finish, then hold the remaining queue.
Resume: continue the paused queue.
Stop square: interrupt the current image or animation job and discard that in-progress result. Use deliberately.

Back to contents ↑

11. Add lip-sync to visible singing scenes

Lip-sync changes the mouth to follow the scene’s assigned song segment. Writing “singing” in a motion prompt alone is not the same as applying audio-driven lip-sync.

Use a forward-facing or gentle three-quarter close-up or medium shot. The lips and jaw must be visible.
Open Edit scene and enable This is a visible singing or performance scene. Save it.
Create and review the base animation first.
Wait until image and animation queues finish.
Choose Lip-sync vocals for one scene. Use Lip-sync all performances in the export panel only when ready to process every marked performance.
Play the result with sound. Check mouth timing, natural closure, and face stability.

A mask, hand, microphone, heavy hair, or deep shadow over the mouth can prevent useful results. A tiny face or fast head turn can also fail. Use a simpler shot rather than repeatedly retrying the same unsuitable clip.

If timing consistently leads or trails the vocal, report the scene and approximate difference. The host handles timing offsets and engine repairs. Lip-sync remains a beta feature; report failures rather than assuming every face will work.

Back to contents ↑

12. Export and download your finished video

Before export, check that the correct song is present, every scene has a rendered animation, desired lip-sync is complete, and no other jobs are running. Check title/outro cards for the orientation you want when those sections are enabled.

Review the scene order and fix rejected or unfinished shots. An approval label alone does not create missing video.
Choose Export 16:9 master for a wide landscape video, or Export 9:16 master for vertical social video.
Wait while the Studio assembles scenes, transitions, title/outro sections, and the original song.
When ready, choose Download master.
Save the file on your device. Check your browser’s Downloads list or your phone’s download/share prompt.
Open the downloaded file and watch it from beginning to end with sound.

Look for missing scenes, black frames, frozen endings, audio gaps, incorrect timing, or clipped titles. In vertical video, make sure the center crop keeps the face and important action visible.

A transition describes how one shot joins the next. “Hard cut” switches immediately; “dissolve” blends briefly; “fade to black” passes through darkness. The final transition behavior depends on host export settings.

Back to contents ↑

13. Finish your turn and return later

Save scene edits using the scene editor’s save control. Finish uploads and check for active or queued jobs.
Record your project name and download important finished exports. Projects and media live on the host, not exclusively on your phone.
If a job is still running, coordinate with the host before handing over control. Releasing access does not mean a job has finished.
Use the session’s release control button when done. Its exact label may vary by build.
When returning, sign in, confirm it is your turn, and select your existing project through the project controls. Check the project name before editing.

A host restart ends your remote session. Sign in again with the current code. If your project is missing, ask the host before creating a duplicate or editing another person’s work.

Back to contents ↑

Troubleshooting: what to do when something goes wrong

The Studio is currently in use: Wait for the other tester to release access. Check Discord before attempting takeover.
Failed to fetch: The browser lost contact with the host. Capture the error, wait about ten seconds, and refresh once. If it persists, ask whether hosting is online. Check the queue before resubmitting.
Forbidden or login appears again: Open the permanent address in a fresh tab and use the current code. A restart may have invalidated your session. If needed, try a private tab for sign-in troubleshooting, not to bypass an occupied session.
A button is unavailable: Check for a missing starting image, unsaved prompt, unchecked scene selection, active queue, or missing animation. Read the scene’s message.
Image or animation failed: Copy the full error, project name, and scene number. Report it in #bug-reports. Do not repeatedly tap Retry or attempt host repairs.
No face detected / mouth does not move: Use a larger, unobstructed face, simpler head motion, and a completed base animation. Confirm the performance setting before lip-sync.
Export will not start: Read the missing-scene or queue message. Finish required animations and lip-sync first. Confirm the song is present.
Page looks stuck: Wait for a status refresh. If necessary, capture the state and refresh once; the host may still be working.
Download is missing: Check the browser’s Downloads list. Confirm export completed and try Download master once more. Report if the finished download is still unavailable.

For any recurring problem, include your device/browser, local time, project, scene, exact action, expected result, actual result, and screenshot. Hide access codes and private information first.

Back to contents ↑

Where to ask for help

#bug-reports: errors, failures, broken controls, or missing downloads.
#general-feedback: generation quality, repeated rejection issues, confusing steps, and usability.
#feature-requests: a capability you want added.
#animation-showcase: creations made with Lockheart Studio, following server rules.
Live Testing/Troubleshooting: voice or screen-sharing help when others are available.

Device and browser:
Date and approximate local time:
Project and scene number:
Button pressed / steps taken:
What I expected:
What happened:
Exact error message:
Does it happen again?
Screenshot or recording (private information hidden):

Guide prepared October 15, 2026. Based on the existing Full Beta User Guide and the later permanent-domain and Single Tester Access instructions. Host settings and later beta changes may affect results.

Back to contents ↑