AI content creation
Use pictures & videos to complete picture and video tasks, and then use voice & music to create dialogue, narration, sound effects and soundtracks.
Image & Video
Image & Video Suitable for quickly completing a picture, video or editing task. The difference from Flows is that this works around a single generation session; Flows connects multiple steps into a process that can be run repeatedly.
Choose the right creation method
- Image generation:Generate images from text or reference material. The prompts focus on the subject, environment, composition, light and style; when using reference materials, clarify what needs to be retained and modified.
- Video generation:Via text or Image generation video. Focus on describing subject movement, shot changes, chronology, and the appearance that needs to remain the same.
- Video editing:Process existing videos, including only loop generation, loop cropping, Frame Interpolation and FlowWall DetailX.
Four functions included in video editing
- Loop generation:Cut off a few frames at the beginning and end of the video, and then use the model to generate about 1.5 seconds of connecting content, allowing the end to naturally connect back to the beginning.
- Loop cutting:Find the two most similar frames within 1 second at the beginning and end of the video, and trim other unnecessary clips so that the remaining clips are connected end to end.
- Frame Interpolation：Increase the video frame rate to make motion pictures smoother.
- FlowWall DetailX：Video super-resolution and frame interpolation are performed simultaneously in one process, improving clarity and making motion smoother.
Use reference material to control results
After uploading reference material, don’t just write “refer to this picture”. It should indicate whether it is a character, costume, setting, style, composition, first or last frame. For example: "Reference picture 1 is only used to maintain the appearance of the characters; reference picture 2 is only used for the scene and lighting." This can reduce the interference between different references.
When using reference materials in Image generation, you should write both retained items and modified items at the same time, for example: "Only replace the cup, keep the characters, desktop, lens and other objects unchanged." When making local modifications, you can first use the annotation tool to point out the location, and then explain the specific changes in the text.
Add reference image
After clicking "Add Reference Image" in "Pictures & Videos", the material selection window will open. Available at My uploads Select an uploaded image, or switch to AI generated Select the generated image; after selecting it, add it as the reference image generated this time.
Set up model and parameters
Select the model first, then set the scale, resolution, quantity and quality. The parameters displayed by different models will be different; if an option does not appear, it means that the current model does not support this setting. When you need to experiment with composition, you can first test with a lower-cost setting, and then increase the resolution or quality after the direction is determined.
Continue processing, downloading and citing
After opening the generated results, continue processing according to the actions provided by the card. Pictures can be used as references to continue generating; videos can use loop generation, loop cropping, Frame Interpolation or FlowWall DetailX. Satisfactory results can be downloaded or passed Reference Continue for the next round of generation.
Voice & Music
enter Voice & Music. Top contains Discover、Text to speech and Audio generation Three tabs, the upper right corner is Create voice。
Discover: Find the right sound
- in Voice Library Filter by language, gender and style, or search for sound names directly.
- Click the preview button to listen and judge based on the audio description whether it is more suitable for character dialogue, narration, short drama or game content.
- Click Use voice, the system will bring the sound into Text to Speech.
- Commonly used sounds can be added Favorites;The sound you created is in My Voices in management.
Text to speech: convert text into speech
- Enter the text that needs to be read aloud; long text is first split according to paragraphs and speakers, so that it can be adjusted and redone separately.
- Click Select voice Select a sound and then select an available audio model.
- adjust Speed、Pitch and Volume;Open when more control is needed Parameters。
- Click Generate. After generation, listen to pauses, emotions and pronunciation before downloading the results.
- The history area on the right can be searched and filtered by sound, model and date; old results can be played or downloaded directly.
Audio generation: a description generates the entire audio
The model used on this page is the same as the Audio generation node on the canvas (Seed Audio 1.0). Unlike Text to speech, which reads a piece of text, here the entire sound scene is handed over to the model: the soundtrack, multi-character dialogue, and environmental sound effects can be generated separately, or they can be written in the same description and generated at once, and the output is a mixed audio track.
- 1The third of the three tabs at the top, the first two being Discovery and Text to speech.
- 2Reference sounds: The plus sign on the left selects from the sound library, and the button on the right uploads your own recordings, up to 3.
- 3Command and Polish: Insert timestamps or timbre references; the Magic Wand turns a brief into a playable script.
- 4One-click translation to English, clear and rewrite, the right side is the real-time word count, the upper limit is 3000.
- 5Expand the speaking speed, tone, and volume, and confirm the estimated points to generate.
- 6The official templates are divided into four categories; History includes the audio generated by this page.
How to write the prompt
When there is a lot of content, it is most stable to write it in three paragraphs: first explain who will speak, then explain what the occasion is, what sounds are in the background, and finally write how to perform each sentence in the order of occurrence. Let's use "Cross-Country in a Rented Van" in the template on the right to open it up - it is a two-person podcast. Click on the template to see the complete original text on the left.
| Section | What goes in it | From the template |
|---|---|---|
| 1. Role and timbre | The gender, age group, speaking speed, accent and personality background of each character, one person and one line | Host A is a woman in her early thirties. Warm and steady, mid-paced American English… |
| 2. Occasion and background sound | In what space was it recorded? What should and should not be in the background? | A small clean recording booth with no background music. Keep natural breaths, a faint chair creak… |
| 3. Performances in order | Who, how, what to say. Lines are enclosed in quotation marks, and performance descriptions are written outside the quotation marks. | Host B laughs: "Not by myself. Me and a mattress that slid forward every single time I braked." |
Three lessons:All lines must be in quotation marks, the model will treat words outside quotation marks (laughs, slower, gently) as performance prompts instead of pronouncing them;Please indicate the unwanted sounds, for example, if only narration is required, "no music" is required, if only ambient sound is required, no characters are required;Closing method at the end, like the last sentence of the template "The two chuckled softly and the sound of turning pages ended", it is cleaner than letting it stop on its own.
| What do you want | How to write the prompt |
|---|---|
| Just the human voice | Write out the characters, lines, and tone, and specify that you don’t want music. Suitable for narration, audio books, interviews |
| Just sound effects or ambient sounds | Write around the source of the sound, its distance, its rhythm, and when it comes in and when it exits, not the characters. |
| Just the soundtrack | Write about style, instrumentation, tempo and mood, and indicate that no vocals are required |
Choose who speaks
Click the plus sign above the prompt word box to open the sound library and select one, up to three. When the mouse is parked on the plus sign, it will show how many have been added. When the number is full, the button will turn gray. The sounds that have been added are displayed as a row of avatars. You can stop and listen to them. The cross in the upper right corner can be removed. If you want to use your own voice, click the upload button next to it. After adding it, enter it in the text @ Quote it to specify which voice should be used to pronounce a certain line.
Timestamps and voice references
When you need to make a certain line or sound effect fall within a fixed number of seconds, click the command button in the lower left corner of the prompt box:Add timestamp After filling in the start and end seconds, a shape like this will be inserted at the cursor [2.0s:5.0s] mark;Reference audio It inserts a reference to the reference tone. When no tone is added, it is gray.
Let the model expand your draft
The magic wand next to the command button isPolish prompt: The model will fill in the prompt words more completely and directly replace the main text - the creative brief will be filled in into a performable script, and the already written lines will only be filled in with performance instructions. Stopping the mouse will first display the approximate number of points it will cost this time. After polishing, the button becomes undoable, and you can click it again to restore it to the original.
Speed, pitch and volume
Point Parameters Expand the three sliders, the defaults are 1, 0, and 1.
- 1Speech rate throughout the paragraph.
- 2The pitch of the entire paragraph is expressed in semitones.
- 3The loudness of the entire section.
Start from a template
Right side Templates Official examples are given according to four categories. You can click Play to listen to each item first. Click on the title and fill in the entire prompt word on the left side. You can use it by changing a few words. When you don't know where to start, it's much faster to start with the closest category than to start with an empty box.
- 1Four categories: pure text generation, reference tone, stuck time point, and multilingual.
Generate and history
Click after confirming the estimated points Generate. The maximum length of a single audio is 120 seconds. The generated audios are all in the upper right corner. History Here you can listen, compare and download, and you can also find it in Assets.
Music: generate from lyrics and genre
Music and the Audio generation above are two different things: Audio generation is a "whole sound scene", with vocals, sound effects, and soundtrack mixed in one track; the music page only does pure music or songs with lyrics, and the output is a complete song.
- Fill in the lyrics. When you don’t need fixed lyrics, you can turn off this option and let the model perform as described. What you get is pure music or humming.
- Write a description of the music: genre, instrumentation, tempo and mood, such as light acoustic guitar playing, medium tempo, suitable for a travel vlog intro.
- After selecting the model, it is generated. The completed song can be listened to, downloaded, or used as material in Flows and connected to the Video composition node.
Create voice and sound management
You can use it when you need to maintain the same character's voice for a long time Create voice Created and in My Voices in management. Prepare clean, clear sound samples with no background music before creating; only upload sounds that are your own or have been explicitly authorized to do so.