Last verified: July 19, 2026. Dream Track began as a limited YouTube experiment in 2023. Do not assume its original artist-voice catalog or account availability describes the current Shorts product.
Dream Track was an experiment, not a universal feature
YouTube introduced Dream Track as a small experiment for selected creators. Participants could prompt a 30-second soundtrack using authorized artist voices. It was powered by an earlier Google DeepMind Lyria generation model.
That history is useful, but old instructions such as “open every YouTube account and select Dream Track” are misleading. Feature names, markets, models, and eligibility have continued to change.
Primary historical source: YouTube's early Dream Track announcement.
What is current now?
There are three separate things creators often mix together:
| Product/workflow | Where it runs | What to verify |
|---|---|---|
| Original Dream Track experiment | Selected YouTube Shorts accounts | Current account eligibility and whether the feature still appears |
| Newer YouTube AI creation/remix tools | YouTube/YouTube Create in supported markets | Model, source attribution, rollout status, and creator controls |
| Lyria 3 Gemini API | A developer's application | API key, model availability, safety rules, cost, and output rights |
YouTube's 2025 creation update, for example, described a Speech to Song remix feature powered by Lyria 2—not a blanket rollout of the original Dream Track artist-voice flow. YouTube also described content labels and SynthID for those AI creations.
Source: YouTube's generative creation tools update.
A dependable Shorts workflow today
You do not need to rely on a particular YouTube experiment being present:
- Define the Short's duration and energy changes.
- Generate a 30-second instrumental draft with a clearly labeled model.
- Listen under the voiceover, not only on headphones by itself.
- Create a variation if the hook competes with speech.
- Export the original and preserve provider/model records.
- Add the track in your video editor, then upload to Shorts.
- Review YouTube's current synthetic-content, copyright, and music policies before publishing.
Generate a 30-second music direction
Prompt around the edit
A useful prompt specifies the timeline:
30-second upbeat electronic instrumental for a product reveal. 0:00–0:05 minimal pulse, 0:05–0:18 rising percussion under voiceover, 0:18–0:24 bright hook for the reveal, 0:24–0:30 clean button ending. No vocals.
For a talking Short, request sparse mids and no vocals. For a montage, ask for stronger rhythmic landmarks at cut points. For looping playback, request an ending compatible with the opening texture.
Artist voices require explicit authorization
The original Dream Track experiment used participating artists who had authorized the experience. That does not make imitation of any artist generally permitted. In Lyria 3's current API documentation, prompts requesting specific artist voices or copyrighted lyrics can be blocked by safety filters.
Use descriptive musical attributes—register, texture, tempo, instrumentation, era, and mood—instead of a living artist's name.
Provenance and rights
Google says official Lyria 3 audio contains SynthID. YouTube has also described watermarks and content labels for its supported AI tools. However:
- a watermark is not a commercial-use license;
- output from a third-party model is not automatically a Lyria/SynthID output;
- platform policies can require disclosure even when an audio watermark exists;
- uploaded samples, lyrics, and voice likeness create separate rights questions.
Save the model, prompt, source assets, generation date, and applicable terms for each final soundtrack.
Before publishing
Check the live YouTube app and official Help/Blog pages for your region. If a creation feature is not present, do not follow an old workaround or claim access you do not have. A manually imported, properly documented soundtrack is a more reliable production workflow than depending on an experimental menu item.
Start with one free generation and review current plan options.